Industrial and commercial energy storage power supply system for flexible distribution network

Abstract

The invention relates to the field of energy storage and power supply, in particular to an industrial and commercial energy storage and power supply system for a flexible distribution network, which comprises the following steps: acquiring multi-source data through an Internet of Things sensing device, grading according to load reliability, and configuring an independent power supply loop and a data acquisition unit; determining an energy storage access point by using algorithms such as deep learning and Kalman filtering, and dynamically adjusting the pre-storage capacity; a multi-mode interconnection topology is constructed through equipment such as a flexible intelligent switch, and energy storage cooperative scheduling is realized in combination with reinforcement learning and droop control; a power market is linked by means of a multi-time scale optimization model and a shadow price mechanism, and an economical and reliable scheduling scheme is generated; and finally, fault self-healing is realized through graded early warning and flexible equipment, so that the flexibility, reliability and economy of the distribution network are improved.

Description

Technical Field

[0001] The present invention relates to the field of energy storage and power supply, and in particular to an industrial and commercial energy storage and power supply system for a flexible distribution network. Background Art

[0002] With the intelligent transformation of industry and commerce, industrial and commercial distribution networks face multiple technical challenges: highly volatile loads (instantaneous power fluctuations can reach 30%-50%) place strict demands on real-time power balance, and graded reliability requirements (such as the allowed power outage time for first-level loads is less than 50ms) urgently require differentiated power supply strategies. The proportion of distributed photovoltaic / energy storage installed capacity has increased to 40%-60%, but the intermittent nature of renewable energy has led to a "waste rate" of 15%-20%. In addition, the peak-to-valley electricity price gap has expanded to more than 3 times, which has spawned economic scheduling needs. All of these pose challenges to the flexibility, reliability and economy of traditional distribution networks.

[0003] However, existing data fusion technologies only focus on electrical quantity collection, without integrating meteorological, equipment status, and electricity price signals, resulting in load forecast errors exceeding 10%. Flexible regulation relies on manual intervention, with fault recovery times of several minutes and a lack of pulse support for instantaneous impact loads. Multi-objective optimization focuses on single economic efficiency, ignoring reliability (such as primary load power supply continuity) and environmental protection (renewable energy absorption rate). Distributed power sources, energy storage, and load terminals work inefficiently, and when there is excess photovoltaic power, the energy storage charging power utilization rate is less than 70%.

[0004] To this end, there is an urgent need for an industrial and commercial energy storage and power supply system for flexible distribution networks. By deeply integrating multi-source data, building a flexible interconnection topology, and innovating intelligent optimization algorithms, the energy storage and power supply method can break through the technical bottlenecks of source-load-storage coordinated scheduling, fault self-healing, and multi-objective balance to meet the complex energy supply needs of industrial and commercial scenarios. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an industrial and commercial energy storage and power supply system for a flexible distribution network. By deeply integrating multi-source data, building a flexible interconnection topology, and innovating an intelligent optimization algorithm for energy storage and power supply methods, the system can meet the complex energy supply needs of industrial and commercial scenarios.

[0006] To achieve the above objectives, one of the present inventions is to provide an industrial and commercial energy storage power supply system for a flexible distribution network, comprising the following modules:

[0007] Multi-source data acquisition module: IoT sensors deployed at key nodes of the distribution network collect real-time distribution network operation data, equipment status data, environmental and meteorological data, real-time electricity market price signals, and three-phase current values ​​of the substation. Based on the reliability requirements of industrial and commercial loads, loads are divided into primary, secondary, and tertiary levels, and independent power supply circuits and data acquisition units are configured for different levels of loads.

[0008] Energy storage pre-configuration module: Utilizes deep learning models and Kalman filtering algorithms, combined with the three-phase current values ​​and voltage over-limit responsibility quantification algorithm in the substation area, to determine the priority energy storage access point. It calculates the initial energy storage capacity of the access point based on the particle swarm optimization algorithm and dynamically adjusts the energy storage pre-reserve based on real-time data.

[0009] Flexible interconnection control module: This module builds a flexible interconnection topology in the power supply network, including star mode, constellation mode, and pulse mode. It generates an adaptive power allocation strategy through a reinforcement learning algorithm, and coordinates the charging and discharging power of energy storage units in combination with a droop control algorithm to achieve multi-energy flow coordinated control of distributed power sources, energy storage units, and load terminals.

[0010] Multi-objective optimization module: This module establishes a multi-timescale optimization model, generates energy storage dispatch constraints based on a shadow price model, and uses electricity market transaction data to update the optimization objective function that includes economic, reliability, and environmental goals. It also uses a genetic algorithm to jointly optimize energy storage charging and discharging plans, load transfer plans, and power supply plans.

[0011] Self-healing control module: Establishes hierarchical warning thresholds. Based on equipment status evaluation and risk warning results, it dynamically adjusts the power supply path and power distribution plan through IoT sensors and flexible interconnected devices at key nodes of the distribution network to achieve automatic fault detection, isolation, power outage transfer and voltage regulation, thereby improving the system's self-healing capability and reliability.

[0012] The second aspect of the present invention is to provide an industrial and commercial energy storage power supply method for a flexible distribution network, which is used to implement an industrial and commercial energy storage power supply system for a flexible distribution network, comprising the following steps:

[0013] S100: IoT sensors deployed at key nodes of the distribution network collect real-time operational data, equipment status, and topology data within the distribution network. This data is combined with environmental and meteorological data, three-phase current values ​​in the distribution area, and real-time electricity price signals from the power market. The system then categorizes loads based on the reliability requirements of industrial and commercial loads, configuring independent power supply circuits and data acquisition units for each load level.

[0014] S200: Utilizes a deep learning model and Kalman filter algorithm, combined with the detected three-phase current values ​​in the substation area and a voltage over-limit responsibility quantification algorithm to determine the energy storage priority access point and dynamically adjust the energy storage reserve based on real-time data;

[0015] S300: Dynamically constructs a flexible interconnected topology in the power supply network, including star mode, constellation mode, and pulse mode. Based on the constructed flexible interconnected topology and the generated adaptive power allocation strategy, combined with the droop control algorithm, it coordinates the charging and discharging power of each energy storage unit to achieve coordinated control of multiple energy flows.

[0016] S400: Establish a multi-timescale optimization model; generate energy storage dispatch constraints based on the shadow price model, and update the optimization objective function in conjunction with power market transaction data;

[0017] S500: Builds a multi-objective optimization model that includes economic, reliability, and environmental goals, establishes graded warning thresholds, and dynamically adjusts power supply paths and power distribution plans based on equipment status evaluation and risk warning results, achieving automatic fault detection, isolation, and self-healing control.

[0018] S600: Through IoT sensors at key nodes of the distribution network, it implements functions such as fault isolation, power outage transfer, dynamic load transfer, and voltage regulation, improving the system's self-healing capabilities and reliability.

[0019] The third aspect of the present invention is to provide a computer device and a readable storage medium, characterized in that the computer device includes: a processor and a memory, the memory storing at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to realize an industrial and commercial energy storage power supply system for a flexible distribution network.

[0020] Compared with the prior art, the present invention provides an industrial and commercial energy storage power supply system and method for a flexible distribution network, which has the following beneficial effects:

[0021] 1. This solution uses IoT sensors to collect data from multiple sources, including distribution network operation, weather, and electricity prices. It divides loads into three tiers based on reliability, deploys supercapacitors for tier-one loads to achieve millisecond-level power outage protection. Deep learning and a voltage over-limit responsibility quantification algorithm are used to determine the optimal energy storage access point. Combined with particle swarm optimization, the pre-storage capacity is dynamically adjusted to address load fluctuations and energy intermittency.

[0022] 2. This solution builds a dynamic interconnection model of "star," "constellation," and "pulse," generates an adaptive power allocation strategy based on a reinforcement learning algorithm, and coordinates the charging and discharging of energy storage units with droop control. This achieves millisecond-level fault isolation and power transfer, minute-level dynamic load transfer, and pulsed power support for instantaneous impact loads, thereby improving distribution network flexibility.

[0023] 3. This solution establishes a multi-objective model encompassing economics, reliability, and environmental protection, employing a genetic algorithm to jointly optimize energy storage scheduling and load transfer. It automatically triggers fault detection and self-healing control using three-level warning thresholds (yellow / orange / red), ensuring an annual power outage time of less than 100ms for primary loads and a renewable energy consumption rate of ≥80%.

[0024] 4. This solution integrates power market transaction data by introducing a shadow price model to achieve minute-level real-time electricity price response, hour-level energy storage scheduling, and daily strategy optimization, minimizing overall energy costs and increasing demand response benefits.

[0025] This solution controls the load forecast error to ±5% through efficient coordination of "source-grid-load-storage", shortens the fault recovery time to within 0.5 seconds, and increases the photovoltaic energy storage charging utilization rate to more than 90%. While ensuring high reliability of industrial and commercial loads, it reduces energy costs by 20%-30%, significantly improving the flexibility, economy and green level of the distribution network, and is suitable for high-elasticity power grid construction and industrial and commercial energy supply scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0027] Figure 1 This is a block diagram of the overall system architecture of the present invention;

[0028] Figure 2 This is a flow chart of load classification and power supply configuration of the present invention;

[0029] Figure 3 A flowchart of the energy storage preconfiguration process of the present invention;

[0030] Figure 4 The flexible interconnect topology switching logic diagram of the present invention;

[0031] Figure 5 This is a flow chart of the hierarchical early warning and self-healing control of the present invention. DETAILED DESCRIPTION

[0032] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0033] It addresses the challenges of high reliability requirements for industrial and commercial distribution networks, distributed energy consumption, economical scheduling, and rapid self-healing.

[0034] This solution proposes an industrial and commercial energy storage power supply system for flexible distribution networks.

[0035] Through IoT sensors at key nodes of the distribution network, multi-dimensional data such as electrical operation, equipment status, weather, and electricity prices are collected in real time.

[0036] The load reliability requirements are divided into three levels: Level 1 loads are equipped with dual power supplies + supercapacitors (10ms power failure switching), Level 2 loads are equipped with lithium batteries + shared circuits (30 minutes of continuous power supply), and Level 3 loads are connected to the demand response system;

[0037] Using the LSTM model to predict load and the Kalman filter to estimate energy storage status, combined with the voltage over-limit responsibility quantification algorithm, particle swarm optimization is used to determine the optimal energy storage access point (prioritizing high-load density and voltage-sensitive nodes).

[0038] By building three dynamic topologies and generating power allocation strategies through reinforcement learning, combined with droop control to coordinate energy storage charging and discharging, this allows for priority storage of excess photovoltaic energy and coordinated multi-source power supply during peak load periods, improving system flexibility.

[0039] Use the NSGA-II algorithm to optimize economics and generate optimal energy storage scheduling and load transfer plans. Establish a three-level warning system (yellow / orange / red);

[0040] Through the integration of multiple technologies and full-process collaboration, it is significantly superior to traditional solutions in terms of reliability, economy and flexibility of industrial and commercial distribution networks. It is particularly suitable for industrial parks with high load fluctuations and high reliability requirements, and has the value of large-scale promotion and application.

[0041] This solution controls the load forecast error to ±5% through efficient coordination of "source-grid-load-storage", shortens the fault recovery time to within 0.5 seconds, and increases the photovoltaic energy storage charging utilization rate to more than 90%. While ensuring high reliability of industrial and commercial loads, it reduces energy costs by 20%-30%, significantly improving the flexibility, economy and green level of the distribution network, and is suitable for high-elasticity power grid construction and industrial and commercial energy supply scenarios.

[0042] Example 1, as Figure 1-Figure 5 As shown, an industrial and commercial energy storage power supply system for a flexible distribution network provided in an embodiment of the present application is exemplified.

[0043] S100: Through IoT sensors at key nodes in the distribution network, multi-dimensional data such as electrical operation, equipment status, weather, and electricity prices are collected in real time. Load reliability requirements are divided into three levels: Level 1 loads are equipped with dual power supplies + supercapacitors (10ms power failure switching), Level 2 loads are equipped with lithium batteries + shared circuits (30 minutes of continuous power supply), and Level 3 loads are connected to the demand response system. Each level is equipped with an independent data collection unit to generate load profiles, providing a data foundation for precise scheduling.

[0044] S200: Utilizing an LSTM model to predict load and a Kalman filter to estimate energy storage status, combined with a voltage over-limit responsibility quantification algorithm, particle swarm optimization is used to determine the optimal energy storage access point (prioritizing high-load density and voltage-sensitive nodes). The initial capacity is configured as a 15-minute backup for primary load and a 30-minute peak-shaving configuration for secondary load. The reserve capacity is dynamically adjusted in real time based on weather conditions and electricity prices (e.g., increasing the SOC to 90% on rainy days) to balance supply and demand with economic efficiency.

[0045] S300: Builds three dynamic topologies: Star Mode (core energy storage radiates primary loads, with millisecond-level fault isolation), Constellation Mode (distributed energy storage interconnection, cross-station load transfer ≤ 200kW), and Pulse Mode (supercapacitors and lithium batteries in parallel to cope with load surges). It generates power allocation strategies through reinforcement learning and coordinates energy storage charging and discharging with droop control, enabling priority storage for excess PV energy and coordinated multi-source power supply during peak load periods, improving system flexibility.

[0046] S400: Establishes minute-level (real-time electricity price response), hour-level (integrated with production plans), and daily-level (long-term strategy) optimization systems, introducing a shadow price model to generate energy storage scheduling constraints (discharge limits during peak hours and charging limits during off-peak hours). The objective function integrates electricity purchase costs, energy storage losses, and demand response benefits, and dynamically adjusts charging and discharging strategies based on power market data to maximize economic and environmental benefits (e.g., photovoltaic absorption rate ≥ 80%).

[0047] S500: Utilizes the NSGA-II algorithm to optimize economic, reliability, and environmental performance, generating optimal energy storage scheduling and load transfer solutions. A three-level warning system (yellow / orange / red) is established, corresponding to pre-load transfer, non-critical load removal, and rapid fault isolation strategies. Fault signatures are identified using wavelet transforms, and the backup topology is activated after locating the fault within 0.1 seconds. Power supply to primary loads is restored within 1 minute, with an annual power outage duration of less than 100ms.

[0048] The S600 uses fiber-optic longitudinal differential protection (20ms actuation) to precisely isolate faults, while a solid-state circuit breaker with a 50μs tripping time ensures primary load outages last less than 50ms. Dynamic load transfer across three levels balances loads (activates when load factor exceeds 85%). Energy storage units, combined with on-load tap-changing transformers, stabilize the primary load voltage within ±2%. A triple-redundant power supply design ensures a mean fault recovery time of ≤3 minutes and a reliability exceeding 99.99%.

[0049] The step S100 is used for multi-source data collection and load classification configuration, and includes:

[0050] S110:

[0051] S1101: Deploy Internet of Things (IoT) sensors at key nodes in the distribution network to form a three-dimensional data collection network:

[0052] 0.2S-class high-precision current / voltage sensors (sampling frequency 1kHz) are installed on the low-voltage side of the transformer, energy storage access points, and load terminals to collect real-time data on three-phase voltage (Ua / Ub / Uc), current (Ia / Ib / Ic), active power (P), reactive power (Q), frequency (f), and total harmonic distortion (THD). The data is uploaded to the edge computing gateway via the ModbusTCP protocol.

[0053] An infrared temperature sensor (accuracy ±1°C) and a vibration sensor (sensitivity 0.1g) are installed on the transformer body to monitor winding temperature and mechanical vibration signals. A BMS (battery management system) is integrated into the energy storage unit to collect real-time data on the lithium battery / supercapacitor's state of charge (SOC, resolution 0.1%), state of health (SOH, calculated based on the ampere-hour integration method), operating voltage (Vcell), and number of cycles (Ncycle).

[0054] Access the weather service API via the MQTT protocol to obtain real-time light intensity (unit: W / m 2 ), wind speed (m / s), temperature (°C), and humidity (%RH); synchronously access the power market trading platform to obtain real-time electricity price signals (including peak and valley time divisions, such as peak time 18:00-22:00 and valley time 0:00-6:00), demand response subsidy standards (such as peak shaving subsidy of 0.8 yuan / kWh) and real-time grid load rate;

[0055] The calculation formula for collecting three-phase current imbalance through the intelligent fusion terminal (TTU) in the substation area is as follows:

[0056]

[0057] Voltage over-limit events (defined as voltage exceeding the rated value by ±10% for more than 10 minutes) and the geographical coordinates of the over-limit node;

[0058] S1102: The edge computing gateway preprocesses the raw data:

[0059] The 3σ rule is used to filter sensor noise. When the voltage data at a certain moment deviates from the mean by more than 3 times the standard deviation, the data completion mechanism (linear interpolation of the previous / next moment data) is triggered.

[0060] Through the IEEE1588 precision time synchronization protocol, the clock deviation of all sensors in the entire network is ensured to be less than 1μs, ensuring the temporal and spatial consistency of data.

[0061] Convert data from different protocols into a unified OPC UA format, generate standardized data streams containing timestamps, node IDs, and data types, and transmit them to the central management platform via the 5G private network.

[0062] S120:

[0063] S1201: Based on the IEC61000-4-11 power quality standard and user demand research, a three-level load classification system is established:

[0064] Class A load:

[0065] Define loads where power interruption will result in significant economic losses (>1 million yuan / minute) or safety incidents, such as semiconductor wafer production lines, data center IT equipment, and medical precision instruments;

[0066] The power interruption time is required to be less than 50ms, the voltage deviation is ≤±2%, and the frequency deviation is ≤±0.1Hz;

[0067] Configured as dual power supply lines (mains power + distributed power supply) + supercapacitor energy storage module (capacity configured at 1.5 times the rated power, discharge time ≥ 15 minutes), and configured with independent feeder loop (cable cross-section selected according to 200% rated current);

[0068] Secondary load (Class B):

[0069] Define loads where power outages affect the continuity of production processes but where short-term restoration is acceptable, such as automated production lines, precision air-conditioning systems, and elevators in commercial complexes;

[0070] The power interruption time is required to be less than 5 minutes, the voltage deviation is ≤±5%, and the frequency deviation is ≤±0.5Hz;

[0071] Configured as a single power supply line + lithium battery energy storage module (capacity configured at 1.2 times the rated power, discharge time ≥ 30 minutes), connected to a shared feeder loop (supporting fast switching);

[0072] Level 3 load (Class C):

[0073] Define non-critical auxiliary loads for which interruption or delay of power supply is acceptable, such as electric heating equipment, landscape lighting, and non-productive office equipment;

[0074] There is no strict continuous power supply requirement, and the load dispatch response time is supported to be ≤10 minutes;

[0075] Configured as a conventional feeder loop, connected to a demand response management system, and equipped with an intelligent circuit breaker (supporting remote opening and closing);

[0076] S1202: Each load level is equipped with an independent data acquisition unit (DCU), integrating a high-precision energy meter (0.5S level) and a communication module (supporting Modbus RTU / TCP), which uploads characteristic data such as load current waveform, power factor, and power usage duration in real time.

[0077] An independent data acquisition unit (DCU) is configured for each load level, integrating a high-precision energy meter (0.5S level) and a communication module (supporting Modbus RTU / TCP), to upload characteristic data such as load current waveform, power factor, and power usage duration in real time;

[0078] S130:

[0079] S1301: The central management platform builds a 3D visualization interface to display in real time:

[0080] Voltage / current waveforms of each node (time axis accuracy 10ms), SOC distribution heat map of energy storage units, load classification status (real-time proportion of A / B / C class loads);

[0081] Abnormal event alarm module: When the voltage deviation of the first-level load is greater than ±3%, a red warning is triggered; when the power supply of the second-level load is interrupted for more than 1 minute, an orange warning is triggered; when the scheduling response of the third-level load is timed out (>15 minutes), a yellow warning is triggered;

[0082] S1302: Use blockchain technology to store key data (such as load classification results and energy storage status) and verify data integrity through hash algorithm (SHA-256) to prevent human tampering or transmission errors.

[0083] Through this data collection and load classification, a multidimensional database was built that encompasses real-time operating status, equipment characteristics, environmental impacts, and user needs. This database clarified the power supply reliability requirements and characteristics of different load levels, providing fundamental data support for the precise integration and differentiated scheduling of subsequent energy storage systems.

[0084] The S200 is used for optimizing energy storage access points and dynamically adjusting pre-storage capacity. Specifically:

[0085] S210:

[0086] S2101: Build an LSTM deep learning model with an input layer consisting of 24 neurons (load data from the previous 24 hours, real-time meteorological data, and electricity price signals), two hidden layers (each with 128 neurons), and an output layer that provides 15-minute load forecasts for the next 24 hours. Training uses the Adam optimizer (learning rate 0.001), with the root mean square error (RMSE) as the loss function and a target error rate of ≤±5%.

[0087] Model formula: ^L(t)=σ(Wh(t-1)+W x x(t)+b h );

[0088] Where ^L(t) is the predicted load at time t, h(t-1) is the hidden state at the previous moment, x(t) is the current input vector, and σ is the activation function;

[0089] Based on the Kalman filter algorithm, the remaining available capacity (SOC) of the energy storage unit is dynamically estimated, taking into account the battery aging effect (capacity decays by 1% every 100 cycles):

[0090]

[0091] Where ^x(t) is the state estimation value, z(t) is the BMS measured SOC, Q is the process noise covariance,

[0092] R is the measurement noise covariance;

[0093] S2102: Compare the predicted value with the actual value every hour. When RMSE>8%, the model fine-tuning mechanism is automatically triggered to update the LSTM model weights through online learning to ensure dynamic convergence of prediction accuracy.

[0094] S220:

[0095] S2201: Define the influence weight W of the transformer on the voltage exceeding the limit node ij , comprehensively considering the voltage deviation degree and line impedance:

[0096]

[0097] in:

[0098] V ij_dev =|U j_measured -U j_rated |U j_rated (node ​​j voltage deviation per unit value, the threshold is set to ±10%);

[0099] (The line impedance from transformer i to node j, ρ is the wire resistivity, L ij is the line length, S ij is the cross-sectional area of ​​the conductor);

[0100] Prioritize W ij >0.2 and load density The node is used as a candidate access point (A j area of ​​the node).

[0101] S2202: Construct a fitness function with the goal of "maximizing voltage regulation efficiency and minimizing energy storage investment costs":

[0102] in:

[0103] is the voltage regulation efficiency (target ≥80%), C inv is the investment cost of the energy storage system, α is the weight coefficient (taken as 0.6);

[0104] During the particle swarm iteration process, each particle represents a set of access point combinations, and the optimal solution is searched through the speed-position update formula. The iteration termination condition is that the fitness value remains unchanged for 50 consecutive generations;

[0105] S230:

[0106] S2301: Based on the load classification results, the initial energy storage capacity must meet the following requirements:

[0107] Primary load reserve capacity: C A =P A ×t A ×1.2(P A is the rated power of the first-level load, t A = 15 minutes, 1.2 is the safety factor);

[0108] Secondary load backup capacity: C B =P B ×t B ×1.1(P B is the rated power of the first-level load, t B =30 minutes, 1.1 is the margin factor);

[0109] Total initial capacity: C init =C A +C B ;

[0110] S2302: When the weather forecast shows that the next day's sunlight intensity is less than 200W / m 2 (PV output drops by more than 50%), start the energy storage charging program 6 hours in advance and increase the SOC to 90% (the charging power is determined by the fuzzy control algorithm to avoid overcharging);

[0111] During peak hours (electricity price > 1.2 yuan / kWh), the energy storage discharge strategy is triggered. When SOC ≥ 70%, the energy storage is discharged at 80% of the predicted load gap:

[0112] Discharge power P disch =min(P load_gap ×0.8,P disch_max );

[0113] During off-peak hours (electricity price < 0.5 yuan / kWh), the charging strategy is activated, charging at a current of 1C based on the rated capacity of the energy storage (P chg =C rated ×1C);

[0114] When the real-time load fluctuation range is greater than 20%, the pre-reserve capacity is dynamically adjusted through the feedforward control algorithm, and the compensation amount ΔC = K p ΔP+K i∫ΔPbt(PID controller parameter K p =0.8, K i =0.2);

[0115] Based on the three-phase current and voltage over-limit data and load classification results collected by S100, the energy storage access points and pre-storage capacity are determined through prediction models and optimization algorithms. This not only meets the rapid response requirements of the primary load, but also provides key node locations for the construction of the flexible interconnection topology in S300, enabling the energy storage system to be accurately deployed in load-sensitive areas, laying the physical foundation for the coordinated control of multiple energy flows.

[0116] Specifically, the S300 is used for flexible interconnection topology construction and multi-energy flow coordinated control, including:

[0117] S310:

[0118] S3101: With the core energy storage node (capacity ≥ 1MWh, equipped with ultracapacitor + lithium battery hybrid energy storage) as the center, it radiates through low-impedance cables (impedance ≤ 0.01Ω / km) to connect primary load nodes within a radius of 100 meters, forming a star topology.

[0119] Solid-state circuit breakers (SSBs, breaking time ≤ 50 μs, on-resistance ≤ 50 mΩ) are configured between nodes to support millisecond-level fault isolation and power switching.

[0120] Semiconductor manufacturing workshops and financial data centers ensure that primary loads receive priority energy storage support during grid fluctuations, with voltage sags less than 5%;

[0121] S3102: Multiple distributed energy storage nodes (50-500 kWh, primarily lithium batteries) are interconnected via flexible intelligent switches (e.g., power electronic transformers, or PETs) to form a mesh topology, supporting cross-station power flow (maximum transfer capacity 200 kW).

[0122] When the load rate of a certain area is greater than 85%, the load transfer algorithm is started;

[0123] Identify a three-level load list (sorted by interruptibility priority);

[0124] Calculate the remaining capacity of adjacent areas (C res =C rated -C used );

[0125] The transfer path is determined by the optimal power flow algorithm (OPA) (minimizing line loss, objective function min∑I 2 R);

[0126] Send instructions to the smart switch to transfer the load in stages (the amount transferred in each stage is ≤ 50kW, and the interval is 10 seconds);

[0127] S3103: When a sudden increase in load current greater than 150% (duration less than 200ms) is detected, it is determined to be an impact load (such as motor starting or welding machine operation);

[0128] The supercapacitor module responds first (discharges within 5ms) and provides peak power (discharge current ≥ 2C, duration 100ms);

[0129] The lithium battery module starts synchronously and takes on continuous power (discharge current ≤ 1C);

[0130] After the shock is over, the supercapacitor is charged first (2C current), and then the lithium battery is charged (0.5C current);

[0131] The supercapacitor and lithium battery are connected in parallel through a bidirectional DC / DC converter, supporting fast switching mode (switching time ≤ 10ms);

[0132] S320:

[0133] S3201: Using the Q-Learning algorithm to build a "state-action-reward" model:

[0134] This includes the energy storage SOC (5 intervals: <20%, 20%-40%, ..., ≥80%), PV output status (excess / balanced / insufficient), and load level ratio (current proportion of A / B / C class loads), for a total of 100-dimensional state vectors.

[0135] 20 power allocation actions, including energy storage charge / discharge power (5 levels: 0.2C / 0.5C / 1C / 1.5C / 2C), grid power purchase / sale status, and distributed power generation priority adjustment;

[0136] R=α·ηRE+β·η cost +γ·η reli (α=0.4, β=0.3, γ=0.3), where ηRE

[0137] is the renewable energy consumption rate, η cost is the energy cost saving rate, η reli Power supply reliability rate for primary load;

[0138] Generate the optimal strategy table through offline training (100,000 simulations), and obtain action instructions by looking up the table during real-time scheduling;

[0139] S3202: Each energy storage node adopts the virtual synchronous generator (VSG) control strategy, and the droop characteristic equation is:

[0140] f=f n -K p (PPn ); U=U n -K q (QQ n );

[0141] Among them, f n / U n is the rated frequency / voltage, P n / Q n is the rated active / reactive power, K p / K q is the droop coefficient (set according to the node capacity, large energy storage node K p =0.05, small node K p =0.1;

[0142] When the SOC of a node is greater than 95%, the droop curve is automatically adjusted to reduce the output frequency and guide the power flow to the adjacent node; when the SOC is less than 15%, the frequency priority is increased and power is taken from photovoltaic or grid first.

[0143] S330:

[0144] S3301: Control Flow:

[0145] Supercapacitor is charged first (target SOC=90%), charging current I sc =min(2C,P surplus / V sc );

[0146] The remaining power is used to charge the lithium battery (target SOC = 80%), and the charging current I lb =min(0.5C, (P surplus -P sc ) / V lb );

[0147] When there is still surplus, electricity is sold to the grid through flexible switches (anti-islanding protection requirements must be met);

[0148] Hardware guarantee: The supercapacitor adopts a double-layer structure (equivalent series resistance ≤ 10mΩ) and supports high-rate charge and discharge; the lithium battery is equipped with a thermal management system (temperature difference control ≤ 5°C) to avoid overcharging and heating;

[0149] S3302: Control process;

[0150] Supercapacitor releases energy (discharge current ≤ 2C, duration ≤ 2 minutes) and responds quickly to instantaneous loads;

[0151] The lithium battery is discharged synchronously (discharge current ≤ 1C, until SOC ≥ 20%) to provide continuous power;

[0152] Start constellation mode and call for energy storage from adjacent substations or purchase electricity from the grid (giving priority to power sources with lower electricity prices);

[0153] Protection mechanism: When the energy storage SOC is less than 10%, low voltage protection is triggered and non-essential loads are cut off (all third-level loads are cut off, and 50% of the second-level loads are cut off according to priority);

[0154] The energy storage access points identified by S200 serve as the core nodes of the flexible interconnection topology. Combined with the load classification results from S100 (such as the distribution of primary loads), the flexible interconnection between energy storage units, distributed power sources, and load terminals is achieved through dynamic switching between star, constellation, and pulse modes. This topological architecture provides flexible power flow paths for S400's multi-timescale optimization, enabling subsequent market-linked scheduling to be precisely controlled based on real-time network status.

[0155] like Figure 3 As shown, step S400 is used for multi-time-scale optimization and market-linked scheduling, including:

[0156] S410:

[0157] S4101: Track real-time electricity price fluctuations to minimize electricity purchase costs during peak periods;

[0158] Get the latest electricity price data every 5 minutes and identify the current time period (peak / flat / valley);

[0159] Calculate the current energy storage available discharge power P disch_avail =P curr_SOC ×1C;

[0160] If the current time is peak and P disch_avail ≥P load_gap , give priority to using energy storage discharge (discharge power = load gap);

[0161] If the energy storage is insufficient, select external power sources based on electricity prices (photovoltaic > adjacent substation energy storage > grid);

[0162] S4102: Weather forecast (PV / wind power output forecast, accuracy ≥85%), user production plan (e.g., factory three-shift load curve, entered into the system one day in advance), equipment maintenance plan (energy storage charging and discharging prohibited periods);

[0163] With the goal of “minimizing the comprehensive cost within a day”, a mixed integer programming (MIP) model is established:

[0164]

[0165] Constraints:

[0166] Energy storage capacity boundary: C min ≤C(t)≤Cmax ;

[0167] Power boundary: P chg_min ≤P chg (t)≤P chg_max , P disch_min ≤P disch (t)≤P disch_max ;

[0168] Load balance: P pv (t)+P es_disch (t)+P grid (t) = P load (t)+P es_chg (t);

[0169] S4103: Optimize the long-term operating efficiency of energy storage and reduce annual operation and maintenance costs;

[0170] Analyze 30 days of historical operating data, calculate the number of charge and discharge cycles of each energy storage node, and trigger capacity upgrade recommendations for nodes that exceed 80% of their design life;

[0171] Adjust load classification thresholds (e.g., increase the upper limit of secondary load capacity by 20%) based on seasonal load characteristics (e.g., air conditioning load proportion increases by 30% in summer);

[0172] Assess the feasibility of expanding distributed power generation (e.g., adding new photovoltaic capacity, calculated based on the payback period model: payback period = initial investment / (annual power generation income + subsidies));

[0173] S420:

[0174] S4201: Introducing the economic shadow price to measure the marginal value of energy storage scheduling:

[0175] SP disch =E peak -C disch_loss =1.5 yuan / kWh-0.15 yuan / kWh=1.35 yuan / kWh;

[0176] Shadow price of charging during valley period: SP chg =E valley +C chg_loss = 0.4 yuan / kWh + 0.1 yuan / kWh = 0.5 yuan / kWh;

[0177] When SP disch >SP chg When charging, it is preferred to discharge during peak hours and charge during valley hours.

[0178] S4202: The lower limit of discharge power during peak hours = predicted load shortfall × 80% to ensure demand response effectiveness; the upper limit of charging power during valley hours = energy storage rated capacity × 1C to avoid overcharging;

[0179] At the end of each day, the energy storage SOC must be ≥30% (to cope with the next morning peak), and can be increased to 60% in special scenarios (such as typhoon warnings);

[0180] S430:

[0181] S4301: When participating in the peak load shifting of the power grid, the income is R DR =∑R DR_cut ×Subsidy standard+R DR_shift ×(peak price - valley price)), where R DR_cut is the interruption load, R DR_shift To transfer load;

[0182] S4302: Receive real-time grid load rate, reserve capacity, and spinning reserve price through the API, and dynamically adjust energy storage dispatch strategies: When grid reserve capacity is less than 10%, prioritize energy storage discharge (even during normal hours) to generate ancillary service revenue.

[0183] The flexible interconnected topology and real-time power allocation strategy built by the S300 system provide a dynamically adjustable physical model for the multi-timescale optimization of the S400 system. By integrating electricity price signals from the S100 system and energy storage reserve data from the S200 system, minute-by-minute real-time optimization can rapidly respond to market price fluctuations. Hourly scheduling optimization can integrate the topology to formulate energy storage charging and discharging plans, thus providing an initial scheduling solution for the multi-objective optimization of the S500 system, including economic and reliability constraints.

[0184] like Figure 4 As shown, step S500 is used for multi-objective optimization and hierarchical self-healing control, including:

[0185] S510:

[0186] S5101: Use the non-dominated sorting genetic algorithm (NSGA-II) to optimize the following objectives:

[0187] Economic objectives (f1):

[0188] Includes the cost of purchasing electricity from the power grid and the cost of energy storage operation and maintenance, with the unit operation and maintenance fee being 0.01 yuan / kWh;

[0189] Reliability target (f2): T interrupt The annual interruption time of the first-level load is ≥99.99%;

[0190] Environmental protection goals (f3): Renewable energy consumption rate, target ≥80%;

[0191] S5102: Generate 100 random solutions (each solution includes energy storage charging and discharging plan, load transfer plan, and power supply path selection);

[0192] The solutions are stratified according to the three objectives, with the first tier being the Pareto optimal solution;

[0193] Simulated binary crossover (SBX) and polynomial mutation are used to maintain population diversity;

[0194] Combined with the crowding calculation, the excellent individuals are retained to enter the next generation, and the number of iterations is 100 generations;

[0195] S520:

[0196] S5201: Level 3 warning threshold setting:

[0197] Yellow Level:

[0198] Triggered when the load rate is 70%-80% or SOC is 20%-30%, triggering response measures:

[0199] Start the third-level load pre-transfer program (notifying users to prepare for an outage) and send a power support request to adjacent substations (priority: energy storage > photovoltaic);

[0200] Make the load rate less than 70% or SOC greater than 30% within 30 minutes;

[0201] Orange Level:

[0202] When the load rate is 80%-90% or the SOC is 10%-20%, the response measures are triggered:

[0203] Automatically cut off 50% of the third-level loads (according to the interruptible priority), control the lithium battery module to increase the discharge power to 1.2C (not exceeding the safety threshold), and trigger the hot standby of the first-level load backup power supply;

[0204] Make the load rate less than 80% or SOC greater than 20% within 5 minutes;

[0205] Red Level:

[0206] When the load rate is ≥90%, the equipment temperature is greater than 120℃, or the current suddenly changes by more than 200%, the trigger response measures are as follows:

[0207] Locate the fault point within 0.1 seconds and disconnect the fault switch within 50ms;

[0208] Activate the backup constellation topology and switch to the adjacent area for power supply;

[0209] Send a fault ticket with GPS location information to the operation and maintenance personnel (accuracy ≤ 10 meters);

[0210] Restore power to primary loads within 1 minute and repair faults within 10 minutes;

[0211] S5202: Use wavelet transform to analyze current signals, extract fault characteristics (such as sudden changes in zero-sequence current and sudden increases in harmonic content), and establish a fault type identification library (including 20 fault modes such as short circuit, overvoltage, and equipment overheating). The identification accuracy rate is ≥ 95%;

[0212] S530:

[0213] S5301: Pre-stored 200+ fault handling policies, for example:

[0214] Open the faulty feeder switch F10, close the tie switch S15, and enable the energy storage power supply in the adjacent substation;

[0215] Immediately stop charging and discharging, start forced air cooling (target: reduce the temperature to below 80°C within 10 minutes), and switch to the backup energy storage module at the same time;

[0216] S5302: After each self-healing solution is generated, simulation verification is performed on the digital twin platform:

[0217] Use the Newton-Raphson method (iteration accuracy 1e-6) to verify whether the voltage distribution meets the requirements (first-level load node voltage deviation ≤ ±2%) and use the eigenvalue method to determine whether the system has the risk of low-frequency oscillation. If the risk level is greater than level 3, automatically adjust the power supply path.

[0218] The multi-timescale optimization results generated by S400 (such as energy storage charging and discharging plans and load shifting schemes) serve as the initial input for the multi-objective optimization in S500. Combined with the equipment status data from S100 and the real-time topology information from S300, the NSGA-II algorithm achieves a global balance between economy, reliability, and environmental protection. A hierarchical early warning mechanism relies on energy storage SOC monitoring by S200 and fault signature identification by S300, ensuring that when an anomaly is detected (such as load overload or equipment overheating), the self-healing control process in S600 is quickly triggered.

[0219] Step S600 is used to enhance self-healing capability and improve reliability, and includes:

[0220] S610:

[0221] S6101: Using optical fiber longitudinal differential protection device (operation time ≤ 20ms), by comparing the current phase and amplitude at both ends of the line, the fault section can be accurately located (positioning accuracy ≤ 50 meters). The calculation formula is: When the phase difference is greater than and the amplitude difference is greater than 30%, it is determined to be an internal fault;

[0222] S6102: Level 1 load transfer: After a fault occurs, the solid-state circuit breaker disconnects the fault side within 50μs, and the backup power supply (supercapacitor) is immediately put into use, ensuring that the power outage time is less than 50ms;

[0223] Secondary load transfer: Switches to the backup path through a flexible intelligent switch, with a transfer time of less than 200ms, during which the lithium battery module continues to supply power;

[0224] Level 3 load transfer: temporarily interrupted during fault repair, and restarted according to priority after power supply is restored;

[0225] S6103: Important nodes are equipped with triple-redundant power supplies: mains power + distributed power supply + energy storage module. If any two power sources fail, the third one automatically takes over, ensuring power supply reliability ≥ 99.999%;

[0226] S620:

[0227] S6201: Based on the load factor of the substation When >, start the transfer;

[0228] Priority sorting: tertiary load > non-critical part of secondary load > critical part of secondary load (transferred only in extreme cases);

[0229] Transfer amount calculation: ΔP=min(P load -0.85S rated ,∑C res_neighbor );

[0230] Path optimization: Use the Dijkstra algorithm to select the shortest path (taking into account line impedance and switch status) to avoid circuitous power supply;

[0231] S6202: Reactive power compensation:

[0232] The energy storage unit inverter supports ±50kVar reactive power regulation. When the node voltage deviation is greater than ±5%, the reactive power output (Q adj =k u ·(U meas -U rated ), k u =10KVar / );

[0233] On-load tap changer coordination: Linked with the transformer on-load tap changer (OLTC), OLTC adjustment is initiated when the voltage deviation is greater than ±8% (adjustment level ±1 each time, adjustment time 20 seconds) to ensure that the voltage is stable within the range of ±2%;

[0234] S6203: System Average Interruption Duration (SAIDI): Level 1 load <100ms / year, Level 2 load <5 minutes / year, Level 3 load <30 minutes / year;

[0235] Voltage qualification rate: first-level load ≥99.99%, second-level load ≥99.95%, third-level load ≥99.5%;

[0236] Load transfer success rate: ≥98% (single transfer time ≤2 minutes, automatic retry 3 times in case of failure);

[0237] S6204: Reliability test plan:

[0238] Simulate single-phase grounding, three-phase short circuit and other faults on the digital twin platform to verify self-healing time and power supply restoration effect;

[0239] Gradually increase the primary load to 150% of the rated power to test the energy storage support capacity and voltage sag amplitude;

[0240] The energy storage unit charging and discharging efficiency and sensor reliability are tested within the temperature range of -20℃ to 50℃.

[0241] Experimental example:

[0242] Scenario: A 300-mu industrial park, including semiconductor production lines (level 1 load, 2000kW), automated production lines (level 2 load, 3000kW), and auxiliary equipment (level 3 load, 1000kW), is equipped with 1MWp photovoltaic power + 2MWh lithium batteries + 500kWh supercapacitors.

[0243] Objective: To verify the effectiveness of the solution in core aspects such as load forecasting, energy storage scheduling, and fault self-healing, and to compare the performance differences between the traditional solution (fixed topology + manual scheduling) and this solution.

[0244] Experimental Step 1:

[0245] Sensor deployment:

[0246] Install 120 IoT sensors (80 current / voltage sensors, 30 temperature sensors, and 10 meteorological sensors) covering transformers, energy storage cabinets, and load terminals;

[0247] Access to the grid's real-time electricity prices (1.5 yuan / kWh during peak hours and 0.5 yuan / kWh during off-peak hours) and weather forecast API;

[0248] Level 1 load: semiconductor production line, equipped with supercapacitors (15 minutes of backup capacity);

[0249] Secondary load: automated production line, connected to the lithium battery shared circuit;

[0250] Level 3 loads: auxiliary equipment, incorporated into the demand response system;

[0251] Data was collected for 7 consecutive days to verify sensor accuracy (voltage error ±0.3%, current error ±0.5%) and time synchronization error <2μs;

[0252] Experimental Step 2:

[0253] The load forecasting algorithm is trained using the LSTM model, which inputs historical load and meteorological data and outputs a 24-hour forecast curve (with a step length of 15 minutes). The Kalman filter is used to estimate the energy storage SOC with an error of less than 1%.

[0254] Then, using a voltage over-limit responsibility quantification algorithm, three high-load nodes were identified as energy storage access points (load density 65kW / hectare, voltage over-limit weight > 0.25). The initial energy storage capacity was configured as follows: 300kWh for primary load, 600kWh for secondary load, for a total of 900kWh.

[0255] Simulating rainy weather (PV output drops by 40%), the energy storage SOC is increased from 70% to 90% 6 hours in advance;

[0256] During peak hours, energy storage discharge is triggered, and the SOC drops from 80% to 50%, meeting the load gap of 80%;

[0257] Experimental Step 3:

[0258] In constellation mode, three energy storage nodes are interconnected through flexible intelligent switches;

[0259] Shock load test: Starting pulse mode, supercapacitor + lithium battery are discharged in parallel, and the voltage drop when the motor starts is reduced from 15% to 3% of the traditional solution;

[0260] When there is excess PV power (output 1200kW, load 800kW): supercapacitors are charged at 2C (300kW), lithium batteries are charged at 0.5C (400kW), and the remaining 100kW is sold;

[0261] During peak load (demand 6000kW, local power supply 5000kW): energy storage discharge 1000kW (supercapacitor 200kW + lithium battery 800kW), and adjacent substations transfer 500kW;

[0262] Experimental Step 4:

[0263] During the peak period (19:00-20:00), the electricity price is detected to be 1.5 yuan / kWh, and energy storage discharge (1000kW) is used first, reducing the electricity purchase cost by 1500 yuan / hour;

[0264] At the same time, combined with the factory's three-shift load curve, a storage charging and discharging plan for the next day is generated (charging 1500kWh during off-peak hours and discharging 1200kWh during peak hours).

[0265] The shadow price of discharge during peak hours is calculated to be 1.35 yuan / kWh, which is higher than the 0.5 yuan / kWh of charging during off-peak hours. This optimizes the charging and discharging strategy and increases revenue by 20%.

[0266] Experimental Step 5:

[0267] Launch the NSGA-II algorithm to generate a Pareto optimal solution (economic efficiency improved by 18%, reliability of 99.99%, and photovoltaic absorption rate of 85%);

[0268] Yellow warning (load rate 75%): Automatically pre-transfer 200kW level 3 load;

[0269] Red warning (feeder short circuit): locate the fault in 0.1 seconds, disconnect the switch in 50ms, and restore power supply to the primary load within 1 minute;

[0270] Simulating a three-phase short-circuit fault, the traditional solution's recovery time is 12 minutes, while this solution only takes 3 minutes, and the first-level load power-off time is less than 50ms;

[0271] The experimental data table is as follows:

[0272] index Traditional solutions This program Improvement Load forecast error (RMSE) 12％ 4.5％ 62.5％ First-level load power outage time 1500ms 45ms 97％ Photovoltaic absorption rate 65％ 88％ 35.4％ Peak-hour electricity purchase cost 2500 yuan / day 1800 yuan / day 28％ Fault recovery time 12 minutes 3 minutes 75％ Voltage qualification rate (first level load) 99.5％ 99.99％ 0.49％

[0273] Through the integration of multiple technologies and full-process collaboration, this solution is significantly superior to traditional solutions in terms of reliability, economy, and flexibility of industrial and commercial distribution networks. It is particularly suitable for industrial parks with high load fluctuations and high reliability requirements, and has the value of large-scale promotion and application.

[0274] Example 2: An industrial and commercial energy storage power supply system for a flexible distribution network, comprising the following modules;

[0275] Multi-source data acquisition module: IoT sensors deployed at key nodes of the distribution network collect real-time distribution network operation data, equipment status data, environmental and meteorological data, real-time electricity market price signals, and three-phase current values ​​of the substation. Based on the reliability requirements of industrial and commercial loads, loads are divided into primary, secondary, and tertiary levels, and independent power supply circuits and data acquisition units are configured for different levels of loads.

[0276] Energy storage pre-configuration module: Utilizes deep learning models and Kalman filtering algorithms, combined with the three-phase current values ​​and voltage over-limit responsibility quantification algorithm in the substation area, to determine the priority energy storage access point. It calculates the initial energy storage capacity of the access point based on the particle swarm optimization algorithm and dynamically adjusts the energy storage pre-reserve based on real-time data.

[0277] Flexible interconnection control module: This module builds a flexible interconnection topology in the power supply network, including star mode, constellation mode, and pulse mode. It generates an adaptive power allocation strategy through a reinforcement learning algorithm, and coordinates the charging and discharging power of energy storage units in combination with a droop control algorithm to achieve multi-energy flow coordinated control of distributed power sources, energy storage units, and load terminals.

[0278] Multi-objective optimization module: This module establishes a multi-timescale optimization model, generates energy storage dispatch constraints based on a shadow price model, and uses electricity market transaction data to update the optimization objective function that includes economic, reliability, and environmental goals. It also uses a genetic algorithm to jointly optimize energy storage charging and discharging plans, load transfer plans, and power supply plans.

[0279] Self-healing control module: Establishes hierarchical warning thresholds. Based on equipment status evaluation and risk warning results, it dynamically adjusts the power supply path and power distribution plan through IoT sensors and flexible interconnected devices at key nodes of the distribution network to achieve automatic fault detection, isolation, power outage transfer and voltage regulation, thereby improving the system's self-healing capability and reliability.

[0280] The above module is realized through an industrial and commercial energy storage power supply method for a flexible distribution network.

[0281] Embodiment 3: A computer device and a readable storage medium, characterized in that the computer device includes: a processor and a memory, the memory storing at least one instruction, at least one program, a code set or an instruction set, the at least one instruction, the at least one program, the code set or the instruction set being loaded and executed by the processor to realize an industrial and commercial energy storage power supply system for a flexible distribution network.

[0282] While 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 these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An industrial and commercial energy storage power supply system for a flexible distribution network, characterized in that: Includes the following modules: Multi-source data acquisition module: IoT sensors deployed at key nodes of the distribution network collect real-time distribution network operation data, equipment status data, environmental and meteorological data, real-time electricity market price signals, and three-phase current values ​​of the substation. Based on the reliability requirements of industrial and commercial loads, loads are divided into primary, secondary, and tertiary levels, and independent power supply circuits and data acquisition units are configured for different levels of loads. Energy storage pre-configuration module: Utilizes deep learning models and Kalman filtering algorithms, combined with the three-phase current values ​​and voltage over-limit responsibility quantification algorithm in the substation area, to determine the priority energy storage access point. It calculates the initial energy storage capacity of the access point based on the particle swarm optimization algorithm and dynamically adjusts the energy storage pre-reserve based on real-time data. Flexible interconnection control module: This module builds a flexible interconnection topology in the power supply network, including star mode, constellation mode, and pulse mode. It generates an adaptive power allocation strategy through a reinforcement learning algorithm, and coordinates the charging and discharging power of energy storage units in combination with a droop control algorithm to achieve multi-energy flow coordinated control of distributed power sources, energy storage units, and load terminals. Multi-objective optimization module: This module establishes a multi-timescale optimization model, generates energy storage dispatch constraints based on a shadow price model, and uses electricity market transaction data to update the optimization objective function that includes economic, reliability, and environmental goals. It also uses a genetic algorithm to jointly optimize energy storage charging and discharging plans, load transfer plans, and power supply plans. Self-healing control module: Establishes hierarchical warning thresholds. Based on equipment status evaluation and risk warning results, it dynamically adjusts the power supply path and power distribution plan through IoT sensors and flexible interconnected devices at key nodes of the distribution network to achieve automatic fault detection, isolation, power outage transfer and voltage regulation, thereby improving the system's self-healing capability and reliability.

2. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the multi-source data acquisition module: The first-level load is equipped with an independent power supply circuit consisting of dual power lines and supercapacitor energy storage modules, which supports power-off switching within 10ms and voltage deviation is controlled within ±2%; The secondary load is equipped with a lithium battery energy storage module and a shared power supply circuit, supporting a 50ms response and 30 minutes of continuous power supply. The three-level load is connected to the demand response system, and an intelligent circuit breaker is configured to achieve remote opening and closing control.

3. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the energy storage pre-configuration module, the deep learning model uses an LSTM neural network, inputs historical load data, meteorological data, and electricity price signals, and outputs a 15-minute load forecast curve for the next 24 hours, with a prediction error rate of ≤±5%. The Kalman filter algorithm estimates the state of charge (SOC) of the energy storage unit in real time and introduces a battery aging correction factor (derating by 1% every 100 cycles).

4. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the flexible interconnection control module, the star mode radiates the connection of primary loads with the core energy storage node as the center, and configures solid-state circuit breakers to achieve 50μs-level fault isolation; the constellation mode interconnects distributed energy storage nodes through flexible intelligent switches, supports cross-station load transfer, and the transfer capacity is ≤200kW; the pulse mode temporarily activates the supercapacitor and lithium battery in parallel to discharge for impact loads. The supercapacitor provides a high-power pulse for the first 100ms (discharge current ≥2C), and the lithium battery takes on the subsequent continuous power (discharge current ≤1C).

5. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the multi-objective optimization module, multi-time scale optimization includes minute-level real-time optimization (1-15 minutes), hour-level plan optimization (1-24 hours) and daily strategy optimization (1-7 days); The shadow price model calculates the shadow price of discharge during peak hours (real-time electricity price - discharge loss cost) and the shadow price of charging during valley hours (valley electricity price + charging loss cost), generating upper and lower limit constraints for energy storage scheduling power.

6. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the self-healing control module, the graded warning thresholds include yellow warning (load rate 70%-80% or energy storage SOC 20%-30%), orange warning (load rate 80%-90% or energy storage SOC 10%-20%), and red warning (load rate ≥90% or equipment temperature >120°C), which respectively trigger load pre-transfer, non-critical load removal and fault rapid isolation strategies, with a fault recovery time of ≤3 minutes.

7. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the multi-source data acquisition module, the IoT sensing device includes a 0.2S-level current / voltage sensor (sampling frequency 1kHz), an infrared temperature sensor (accuracy ±1°C) and a BMS battery management system. The collected data is filtered out of the abnormal values ​​using the 3σ rule, and the entire network clock is synchronized (deviation <1μs) through the IEEE1588 protocol, and is converted into OPCUA format and uploaded to the central management platform.

8. The industrial and commercial energy storage power supply system for a flexible distribution network according to claim 1, characterized in that: In the energy storage pre-configuration module, the voltage over-limit responsibility quantification algorithm is calculated using the formula: Calculate the transformer's influence weight on the voltage exceeding the limit node, where V ij_dev is the node voltage deviation per unit value L ij is the line impedance; the particle swarm optimization algorithm determines the energy storage access point combination and capacity configuration with the goal of maximizing voltage regulation efficiency (≥80%) and minimizing energy storage investment cost.

9. An industrial and commercial energy storage power supply method for a flexible distribution network, used to implement an industrial and commercial energy storage power supply system for a flexible distribution network as claimed in any one of claims 1 to 8, characterized in that: The following steps are involved: S100: IoT sensors deployed at key nodes of the distribution network collect real-time operational data, equipment status, and topology data within the distribution network. This data is combined with environmental and meteorological data, three-phase current values ​​in the distribution area, and real-time electricity price signals from the power market. The system then categorizes loads based on the reliability requirements of industrial and commercial loads, configuring independent power supply circuits and data acquisition units for each load level. S200: Utilizes a deep learning model and Kalman filter algorithm, combined with the detected three-phase current values ​​in the substation area and a voltage over-limit responsibility quantification algorithm to determine the energy storage priority access point and dynamically adjust the energy storage reserve based on real-time data; S300: Dynamically constructs a flexible interconnected topology in the power supply network, including star mode, constellation mode, and pulse mode. Based on the constructed flexible interconnected topology and the generated adaptive power allocation strategy, combined with the droop control algorithm, it coordinates the charging and discharging power of each energy storage unit to achieve coordinated control of multiple energy flows. S400: Establish a multi-timescale optimization model; generate energy storage dispatch constraints based on the shadow price model, and update the optimization objective function in conjunction with power market transaction data; S500: Builds a multi-objective optimization model that includes economic, reliability, and environmental goals, establishes graded warning thresholds, and dynamically adjusts power supply paths and power distribution plans based on equipment status evaluation and risk warning results, achieving automatic fault detection, isolation, and self-healing control. S600: Through IoT sensors at key nodes of the distribution network, it implements functions such as fault isolation, power outage transfer, dynamic load transfer, and voltage regulation, improving the system's self-healing capabilities and reliability.

10. A computer device and a readable storage medium, characterized in that: The computer device includes: a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement an industrial and commercial energy storage power supply system for a flexible distribution network as described in any one of claims 1 to 8.

Images