A public building grid-connected power distribution system with light storage coordination

Abstract

This invention discloses a photovoltaic-storage collaborative power distribution system for public buildings, belonging to the field of smart grids. To address the problems of insufficient compatibility in retrofitting older buildings and weak robustness to prediction errors in existing technologies, this invention achieves automatic identification and conversion of communication protocols, voltage levels, and data formats of older equipment through a multi-protocol adaptive interface layer, eliminating integration barriers. Furthermore, it provides phased implementation plans through an older building compatibility assessment and progressive upgrade module, reducing the risk and budget pressure of one-time retrofitting. Through a two-stage stochastic optimization algorithm in a robust optimization engine, it simultaneously considers expected cost and risk variance in the objective function, and introduces a safety boundary that adapts to prediction errors, ensuring that the optimization strategy achieves a balance between the most likely and worst-case scenarios, avoiding the performance collapse of traditional point estimation optimization under error fluctuations.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a photovoltaic-storage synergistic power distribution system for public buildings connected to the grid. Background Technology

[0002] With the rapid development of distributed photovoltaic power generation and energy storage technologies, public buildings (such as office buildings, hospitals, schools, and shopping malls) are increasingly equipped with photovoltaic and energy storage systems to achieve goals such as peak shaving and valley filling, reducing electricity costs, and improving power supply reliability. However, the power distribution of public buildings to the grid faces the following prominent problems.

[0003] The existing power distribution system currently has the following technical problems:

[0004] On the one hand, the compatibility of renovating old buildings is insufficient. A significant proportion of public buildings are existing buildings, which generally face practical engineering limitations such as aging power lines, space constraints, budget constraints, and equipment compatibility. Although existing technologies have interface integration, they have not specifically designed renovation schemes for the special constraints of old buildings, resulting in excessive renovation costs, disruption to normal operation during construction, poor synergy stability between new and old equipment, and a lack of flexibility for phased implementation and optimization upgrades. On the other hand, photovoltaic-storage synergy systems rely on load forecasting and photovoltaic power generation forecasting, but in actual applications, the forecasting error can reach 20-30%. Existing technologies focus on improving forecasting accuracy but do not fully address the impact of forecasting errors on system performance. This leads to a sharp decline in the performance of traditional optimization methods under forecasting errors, weak ability to cope with extreme weather, large fluctuations in system economics, and insufficient power supply reliability.

[0005] To address this, we propose a photovoltaic-storage synergistic power distribution system for grid-connected public buildings. Summary of the Invention

[0006] The purpose of this invention is to provide a photovoltaic-storage collaborative power distribution system for public buildings, which solves the problems of insufficient compatibility with old building renovations and weak robustness to prediction errors in current power distribution systems.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic-storage synergistic power distribution system for public buildings, comprising: a prediction error estimation and probabilistic modeling module, used to collect historical data on load prediction errors and photovoltaic power generation prediction errors in real time, establish a joint probability distribution model of errors at multiple time scales, and generate a set of prediction error scenarios for robust optimization; an old building compatibility assessment and progressive upgrade module, including an automated assessment unit, a renovation feasibility report generation unit, a modular hardware platform, and a progressive control algorithm library, used to assess the renovation constraints of the target building, generate phased implementation plans, and support phased upgrades from basic monitoring to intelligent optimization; and a multi-protocol adaptive interface layer, supporting automatic identification and conversion of multiple communication protocols. The system features adaptive matching of multi-level voltages and unified data format processing to achieve seamless integration between legacy equipment and new systems. A robust optimization engine and real-time control actuator, including a two-stage stochastic optimization solver, a sub-brutal optimization engine, an adaptive risk control unit, and a dynamic safety boundary adjustment mechanism, are used to generate and execute optimal power allocation strategies with robust prediction error. The output of the prediction error estimation and probabilistic modeling module is connected to the input of the robust optimization engine. The output of the legacy building compatibility assessment and progressive upgrade module is connected to the configuration parameters of the multi-protocol adaptive interface layer. The output of the multi-protocol adaptive interface layer is connected to the constraints of the robust optimization engine. The output of the robust optimization engine is connected to the control commands of the real-time control actuator.

[0008] Furthermore, the probability modeling method in the prediction error estimation and probability modeling module includes the following steps: S1: Collect the load prediction error sequence and photovoltaic power generation prediction error sequence for the past 30 days, and calculate their mean, variance, skewness, and kurtosis respectively; S2: Construct a joint probability distribution model of the error, and use a Gaussian mixture model to fit the multimodal distribution characteristics of the error. The model parameters are estimated by the expectation-maximization algorithm; S3: Generate a set of prediction error scenarios: randomly select N scenario samples from the fitted probability distribution, and each scenario contains the error value every 15 minutes in the next 24 hours; S4: Assign a probability weight to each scenario, and the weight is proportional to the frequency of the scenario in the historical data.

[0009] Furthermore, the progressive upgrade control algorithm in the old building compatibility assessment and progressive upgrade module includes the following four stages: Stage 1: Monitoring mode, which only collects data on building load, photovoltaic power generation, and energy storage status to establish a baseline load model without executing any control actions; Stage 2: Assist mode, which provides power allocation suggestions based on the baseline model, which are executed after manual confirmation, and the system records the correspondence between manual decisions and actual effects; Stage 3: Semi-automatic mode, which automatically executes power allocation in normal scenarios where the prediction error is less than 15%, and switches to manual intervention mode in extreme scenarios where the prediction error is greater than 15% or the power grid is abnormal; Stage 4: Fully automatic mode, which trains intelligent control strategies based on reinforcement learning algorithms to achieve adaptive optimization control in all scenarios, and locks the algorithm parameters after the system performance reaches the preset target.

[0010] Furthermore, the multi-protocol adaptive interface layer includes: a protocol identification unit, used to monitor the data frame format, baud rate, and parity bit of the communication port in real time, match it with a preset protocol feature library, and automatically identify the current device's communication protocol type; a data conversion engine, used to convert data frames of different protocols into a unified JSON format, with field mapping relationships stored in a configurable mapping table; a voltage adapter, used to detect the device's rated voltage and switch to the matching voltage level through a solid-state relay array, supporting three standards: 220V±10%, 380V±10%, and 10kV±5%; and a plug-and-play interface, using an RJ45 physical interface, supporting hot-swapping, and completing automatic identification and configuration within 30 seconds after device connection.

[0011] Furthermore, the two-stage stochastic optimization algorithm in the robust optimization engine is modeled as follows: Objective function: The optimization objective is to minimize the total cost, which includes the expected cost and risk cost of the first-stage decision. The first-stage decision variables include the energy storage charging and discharging plan and the photovoltaic power output plan. The risk cost is the product of the variance of the second-stage adjustment cost and the risk aversion coefficient. The second-stage adjustment cost includes grid interaction cost, load shedding penalty, and equipment loss cost. The risk aversion coefficient is dynamically adjusted according to the system operating status. Constraints include: (i) Equipment capacity constraint: The output power of the energy storage battery is limited between its allowable minimum power and maximum power, and the charge of the energy storage battery... (ii) Power balance constraint: The sum of the output power of the energy storage battery, the output power of the photovoltaic power generation and the power of the grid interaction is equal to the building load power demand; (iii) Operational safety constraint: The power of the grid interaction does not exceed the preset safety boundary; (iv) Prediction error adaptive safety boundary: The safety boundary of the grid interaction power is adaptively adjusted according to the prediction error random variable to ensure that the limit is not exceeded in extreme scenarios; wherein, the prediction error random variable follows a joint probability distribution, the output power of the energy storage battery is positive when it is discharging and negative when it is charging, the power of the grid interaction is positive when it is drawing power from the grid and negative when it is feeding power to the grid.

[0012] Furthermore, the system also includes a performance monitoring and feedback optimization module, which is used to collect system operation indicators in real time, including power supply reliability indicators, economic indicators, equipment health indicators and prediction accuracy indicators, and establish a multi-objective optimization function based on the above indicators, and continuously optimize the control strategy parameters through online learning algorithms.

[0013] Furthermore, the system supports three operating modes: grid-connected optimization mode, off-grid autonomous mode, and emergency response mode, which automatically switch according to the grid status and system requirements.

[0014] Furthermore, the off-grid autonomous mode includes the following control logic: Off-grid switching detection: if the grid voltage exceeds the rated range by ±10% for three consecutive sampling cycles, off-grid switching is triggered; Critical load identification: based on historical data, the types of loads and power requirements that must be guaranteed within the building are identified; Power allocation priority: the first priority is safety lighting, fire protection systems, and medical equipment; the second priority is office equipment and air conditioning systems; the third priority is non-essential loads; Energy storage SOC protection: when the energy storage SOC is below 20%, the third priority loads are gradually reduced; when the energy storage SOC is below 10%, the second priority loads are reduced.

[0015] Furthermore, the multi-protocol adaptive interface layer also includes a device health status monitoring unit and a lifespan prediction model, wherein: the device health status monitoring unit collects the operating parameters of the access device in real time, including operating temperature, vibration amplitude, communication error rate, and response latency, and determines the device health status level based on preset threshold rules; the lifespan prediction model uses a survival analysis algorithm to predict the remaining lifespan of the device based on the device's historical operating data, environmental parameters, and load curves, and generates an early warning signal and provides maintenance suggestions when the predicted remaining lifespan is lower than a preset threshold; the interface layer dynamically adjusts the communication sampling rate and data verification strength according to the device health status level, and adds data redundancy verification and retransmission mechanisms for devices with low health status levels.

[0016] Furthermore, the scene probability weights in S4 are dynamically adjustable. The adjustment methods include: weight update cycle: the scene weights are updated every 6 hours based on the latest prediction error data; weight adjustment algorithm: a Bayesian update method is used, taking the real-time prediction error observations as new evidence to update the posterior probability of the scene; weight constraint: the weight adjustment range of any scene does not exceed ±50% of the weight of the previous cycle to avoid drastic weight fluctuations; extreme scene enhancement: when the similarity between the current weather conditions and a certain historical extreme scene exceeds 80%, the weight of that scene is temporarily increased to twice the normal weight.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: The photovoltaic-storage collaborative power distribution system for public buildings proposed in this invention addresses the shortcomings of existing power distribution systems, such as insufficient compatibility with retrofitting older buildings and weak robustness to prediction errors. This invention, on the one hand, achieves automatic identification and conversion of communication protocols, voltage levels, and data formats of older equipment through a multi-protocol adaptive interface layer, eliminating integration barriers. On the other hand, it provides phased implementation plans through an older building compatibility assessment and progressive upgrade module, reducing the risk and budget pressure of one-time retrofitting. Furthermore, it establishes a joint probability distribution model of errors across multiple time scales through a prediction error estimation and probability modeling module, generating a weighted set of prediction error scenarios to provide real uncertainty input for optimization. Additionally, the two-stage stochastic optimization algorithm in the robust optimization engine simultaneously considers expected cost and risk variance in the objective function, introducing a safety boundary that adapts to prediction errors, ensuring that the optimization strategy achieves a balance between the most likely and worst-case scenarios, avoiding performance collapse of traditional point estimation optimization under error fluctuations. Attached Figure Description

[0018] Figure 1 This is an overall architecture diagram of the photovoltaic-storage synergistic power distribution system for public buildings according to the present invention; Figure 2 This is a flowchart of the probabilistic modeling method for the photovoltaic-storage synergistic power distribution system of public buildings in this invention; Figure 3This is a block diagram of the internal structure of the multi-protocol adaptive interface layer of the public building grid-connected power distribution system for photovoltaic-storage synergy of the present invention; Figure 4 This is a block diagram of the robust optimization engine and real-time control actuator of the photovoltaic-storage collaborative power distribution system for public buildings according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To address the technical problems of insufficient compatibility with retrofitting older buildings and weak robustness to prediction errors in existing power distribution systems, such as... Figures 1-4 As shown, the following preferred technical solutions are provided:

[0021] A photovoltaic-storage collaborative power distribution system for public buildings includes: a prediction error estimation and probabilistic modeling module, used to collect historical data on load prediction errors and photovoltaic power generation prediction errors in real time, establish a joint probability distribution model of errors across multiple time scales, and generate a set of prediction error scenarios for robust optimization; an old building compatibility assessment and progressive upgrade module, including an automated assessment unit, a renovation feasibility report generation unit, a modular hardware platform, and a progressive control algorithm library, used to assess the renovation constraints of the target building, generate phased implementation plans, and support phased upgrades from basic monitoring to intelligent optimization; and a multi-protocol adaptive interface layer, supporting automatic identification and conversion of multiple communication protocols and adaptive matching of multi-level voltages. Unified processing of power allocation and data formats enables seamless integration of legacy equipment with new systems. A robust optimization engine and real-time control actuator, including a two-stage stochastic optimization solver, a distributed robust optimization engine, an adaptive risk control unit, and a dynamic safety boundary adjustment mechanism, are used to generate and execute optimal power allocation strategies with robust prediction error. The output of the prediction error estimation and probabilistic modeling module is connected to the input of the robust optimization engine. The output of the legacy building compatibility assessment and progressive upgrade module is connected to the configuration parameters of the multi-protocol adaptive interface layer. The output of the multi-protocol adaptive interface layer is connected to the constraints of the robust optimization engine. The output of the robust optimization engine is connected to the control commands of the real-time control actuator.

[0022] The probabilistic modeling method in the prediction error estimation and probabilistic modeling module includes the following steps: S1: Collect the load prediction error sequence and photovoltaic power generation prediction error sequence for the past 30 days, and calculate their mean, variance, skewness, and kurtosis respectively; S2: Construct a joint probability distribution model of the error, and use a Gaussian mixture model to fit the multimodal distribution characteristics of the error. The model parameters are estimated using the expectation-maximization algorithm; S3: Generate a set of prediction error scenarios: randomly select N scenario samples from the fitted probability distribution. Each scenario contains the error value every 15 minutes in the next 24 hours; S4: Assign a probability weight to each scenario, with the weight proportional to the frequency of the scenario in the historical data.

[0023] By using 30 days of historical data and a Gaussian mixture model, the non-Gaussian and multimodal characteristics of the error are captured, so that the prediction error is no longer a single value, but a probability distribution, providing real input for subsequent robust optimization, accurately quantifying prediction uncertainty, and generating a set of scenarios (including probability weights) that allows the optimization engine to weigh between the most likely and worst scenarios, avoiding the strategy failure caused by point estimation errors in traditional methods.

[0024] The progressive upgrade control algorithm in the old building compatibility assessment and progressive upgrade module includes the following four stages: Stage 1: Monitoring mode, which only collects data on building load, photovoltaic power generation, and energy storage status to establish a baseline load model without executing any control actions; Stage 2: Assist mode, which provides power allocation suggestions based on the baseline model, which are executed after manual confirmation, and the system records the correspondence between manual decisions and actual effects; Stage 3: Semi-automatic mode, which automatically executes power allocation in normal scenarios with prediction errors of less than 15%, and switches to manual intervention mode in extreme scenarios with prediction errors greater than 15% or grid anomalies; Stage 4: Fully automatic mode, which trains intelligent control strategies based on reinforcement learning algorithms to achieve adaptive optimization control in all scenarios, and locks the algorithm parameters after the system performance reaches the preset target.

[0025] The four stages allow users to start with pure monitoring and gradually increase the level of automation, avoiding the technical risks, budget pressure and maintenance burden brought about by one-time transformation. The monitoring stage collects real data, the auxiliary stage incorporates human experience, the semi-automatic stage distinguishes between normal and extreme scenarios, and the fully automatic stage locks in the optimal parameters. The whole process fully considers the actual constraints such as aging wiring in old buildings, space limitations, and phased budget.

[0026] The multi-protocol adaptive interface layer includes: a protocol identification unit, used to monitor the data frame format, baud rate, and parity bit of the communication port in real time, match it with a preset protocol feature library, and automatically identify the current device's communication protocol type; a data conversion engine, used to convert data frames of different protocols into a unified JSON format, with field mapping relationships stored in a configurable mapping table; a voltage adapter, used to detect the device's rated voltage and switch to the matching voltage level through a solid-state relay array, supporting three standards: 220V±10%, 380V±10%, and 10kV±5%; and a plug-and-play interface, using an RJ45 physical interface, supporting hot-swapping, and completing automatic identification and configuration within 30 seconds after device connection.

[0027] Eliminating integration barriers for legacy equipment, the system covers three major integration challenges: inconsistent communication protocols, heterogeneous data formats, and mismatched voltage levels. This allows legacy equipment to be seamlessly integrated into new systems. The plug-and-play interface reduces reliance on professional commissioning personnel and shortens deployment time, making it particularly suitable for budget-constrained renovation projects of old buildings.

[0028] The two-stage stochastic optimization algorithm in the robust optimization engine is modeled as follows: Objective function: Where: x is the decision variable for the first stage, representing the energy storage charging and discharging plan and the photovoltaic power output plan; xi is the random variable of prediction error, which follows the joint probability distribution established by S2; The second stage adjustment cost function includes grid interaction costs, load shedding penalties, and equipment loss costs; lambda is the risk aversion coefficient, which is dynamically adjusted according to the system operating status; constraints: (i) equipment capacity constraints: (ii) Power balance constraints: (iii) Operational safety constraints: (iv) Adaptive safety boundary for prediction error: ;in: : Energy storage battery output power (positive value for discharging, negative value for charging); The minimum / maximum power allowed for energy storage is limited by the physical characteristics of the battery; Photovoltaic power generation output power; Power exchanged with the power grid (positive values ​​indicate power drawn from the grid, negative values ​​indicate power fed back to the grid). Building load power requirements; The minimum / maximum state of charge set to protect the battery; The power grid interaction safety boundary is adaptively adjusted based on random error xi to ensure that the limit is not exceeded in extreme scenarios.

[0029] The objective function considers both expected cost and variance (risk) simultaneously. Constraint (iv) introduces a safety boundary that adapts to the prediction error, so that the optimization strategy remains feasible even when the error fluctuates. This model transforms the problem of weak robustness to prediction error into a solvable stochastic optimization problem, providing a theoretically optimal solution and avoiding the performance collapse of traditional deterministic optimization under error.

[0030] Wherein, the prediction error random variable follows a joint probability distribution, the energy storage battery output power is positive when it represents discharging and negative when it represents charging, and the grid interaction power is positive when it represents drawing power from the grid and negative when it represents feeding power to the grid.

[0031] The system also includes a performance monitoring and feedback optimization module, which is used to collect system operation indicators in real time, including power supply reliability indicators, economic indicators, equipment health indicators and prediction accuracy indicators, and establish a multi-objective optimization function based on the above indicators, and continuously optimize the control strategy parameters through online learning algorithms.

[0032] By continuously tracking key indicators such as power supply reliability, economy, and equipment health, and dynamically adjusting control strategies through online learning, the system can adapt to long-term evolutions such as changes in building load and equipment aging. Through feedback optimization, the system can accumulate experience and improve robustness when dealing with prediction errors, and optimize the gradual upgrade path based on actual operating data in the renovation of old buildings.

[0033] The system supports three operating modes: grid-connected optimization mode, off-grid autonomous mode, and emergency response mode, which automatically switch according to the grid status and system requirements.

[0034] The off-grid autonomous mode includes the following control logic: Off-grid switching detection: if the grid voltage exceeds the rated range by ±10% for three consecutive sampling cycles, off-grid switching is triggered; Critical load identification: based on historical data, the types of loads and power requirements that must be guaranteed within the building are identified; Power allocation priority: the first priority is safety lighting, fire protection systems, and medical equipment; the second priority is office equipment and air conditioning systems; the third priority is non-essential loads; Energy storage SOC protection: when the energy storage SOC is below 20%, the third priority loads are gradually reduced; when the energy storage SOC is below 10%, the second priority loads are reduced.

[0035] The multi-protocol adaptive interface layer also includes a device health status monitoring unit and a lifespan prediction model. The device health status monitoring unit collects the operating parameters of the access devices in real time, including operating temperature, vibration amplitude, communication bit error rate, and response latency, and determines the device health status level based on preset threshold rules. The lifespan prediction model uses a survival analysis algorithm to predict the remaining lifespan of the device based on historical operating data, environmental parameters, and load curves. When the predicted remaining lifespan is lower than a preset threshold, it generates an early warning signal and provides maintenance suggestions. The interface layer dynamically adjusts the communication sampling rate and data verification strength according to the device health status level. For devices with low health status levels, it adds data redundancy verification and retransmission mechanisms.

[0036] The scene probability weights in S4 are dynamically adjustable. The adjustment methods include: weight update cycle: the scene weights are updated every 6 hours based on the latest prediction error data; weight adjustment algorithm: a Bayesian update method is used, which uses real-time prediction error observations as new evidence to update the posterior probability of the scene; weight constraint: the weight adjustment of any scene does not exceed ±50% of the weight of the previous cycle to avoid drastic weight fluctuations; extreme scene enhancement: when the similarity between the current weather conditions and a certain historical extreme scene exceeds 80%, the weight of that scene is temporarily increased to twice the normal weight.

[0037] The performance monitoring and feedback optimization module also includes a multi-stakeholder interest coordination algorithm, which includes: fairness constraints: adding a Gini coefficient constraint to the optimization objective function to ensure that the fairness of energy cost allocation among energy users does not exceed a preset threshold (default 0.3); privacy protection mechanism: using differential privacy technology to process the energy consumption data of each stakeholder, adding random noise that satisfies ε-differential privacy during the data aggregation stage to protect commercial privacy; multi-stakeholder negotiation framework: establishing a negotiation mechanism based on game theory, where each stakeholder participates in the optimization process by submitting preference functions, and the system calculates the Nash equilibrium solution as the final allocation scheme; incentive mechanism design: combining economic incentives (electricity fee discounts) and non-economic incentives (environmental contribution ratings) to construct a multi-dimensional incentive system.

[0038] Implementation Scenario 1: Photovoltaic-storage collaborative renovation project of a municipal government office building.

[0039] The office building was built in 2000. The existing power lines were aging, and the communication protocols were mixed (electricity meters used MODBUS RTU, inverters used CAN bus, and lighting controllers used KNX protocol). Furthermore, the power distribution room had limited space, making it impossible to add a large control cabinet. After adopting this system, the implementation process was as follows: Building compatibility assessment: The automated assessment unit scanned 12 power distribution nodes within the building, generating a feasibility report for the renovation. It was recommended to implement the system in three phases: Phase 1: Deploy monitoring mode and a multi-protocol interface layer; Phase 2: Add energy storage equipment and enable auxiliary mode; Phase 3: Upgrade to semi-automatic / fully automatic mode; Multi-protocol adaptive interface layer: Plug-and-play interfaces connect to existing devices, automatically identifying the protocol type and converting it to a unified JSON format within 30 seconds; Voltage adapters detect that the lighting circuit is 220V and the air conditioning circuit is 380V, automatically switching to the corresponding voltage level; Gradual upgrade control: First… In the monthly operation monitoring mode, a baseline load model is established (daily average electricity consumption of 2200kWh, photovoltaic installed capacity of 150kWp). The following month, the system switches to auxiliary mode, suggesting "energy storage charging during the midday peak photovoltaic output and discharging during the evening peak." This is implemented after manual confirmation, recording a saving of approximately 12% in electricity costs. In the third month, the system enters semi-automatic mode, automatically allocating power under normal weather conditions. When the prediction error exceeds 15% on extreme cloudy or rainy days, manual intervention is required. Robust optimization effect: On a certain day, the weather forecast was sunny, but a sudden cloud cover caused a sharp drop in photovoltaic output of 60%. Under traditional deterministic optimization strategies, the grid interaction power exceeded the limit, triggering a circuit breaker trip. This system, by adopting bibliometric robust optimization and adaptive safety boundaries, increased the energy storage discharge power to 80kW in advance, ensuring a smooth transition and providing power supply reliability.

[0040] Implementation Scenario 2: Emergency Response to Off-Grid Autonomous Mode.

[0041] A coastal public building experienced a typhoon-induced fault in its 10kV power grid. The system detected a voltage drop to -15% of the rated value for three consecutive sampling cycles and automatically switched to off-grid autonomous mode. Key load identification: Based on historical data, fire pumps (15kW), emergency lighting (8kW), and data center servers (10kW) were automatically identified as first priority; office computers and air conditioners were second priority; and canteen equipment and landscape lighting were third priority. Power allocation: The energy storage system's initial SOC was 85%, prioritizing the first priority loads (33kW). After the typhoon lasted for two hours, the energy storage SOC decreased... 20%, gradually cut off the third priority load (12kW canteen equipment); when SOC drops to 10%, cut off part of the second priority air conditioning (retain 5kW minimum ventilation); multi-protocol interface health monitoring: when the interface layer detects that the inverter communication bit error rate rises from 0.01% to 0.5%, the health status level is judged to drop from "good" to "caution", the sampling rate is automatically reduced from 1Hz to 0.1Hz and redundant verification is added to avoid erroneous data causing control strategy failure; grid connection restoration: after the grid is restored, the system smoothly switches back to grid-connected optimization mode, and the energy storage is charged at a 0.2C current-limited rate to avoid impact.

[0042] Implementation Scenario 3: Coordination of interests among multiple stakeholders and dynamic update of scenario weights.

[0043] The building houses three tenants (A: restaurant, B: office, C: retail). This system employs a multi-stakeholder benefit coordination algorithm: Fairness constraint: A Gini coefficient ≤ 0.3 is added to the optimization objective to prevent any tenant from bearing excessively high electricity bills. In a certain month, due to high photovoltaic power generation, the system distributes the energy-saving benefits according to contribution: Tenant A, with 2000kWh of rooftop photovoltaic power generation, receives a 15% discount on electricity bills; Tenant B, without photovoltaic power but with energy storage installed in its distribution room, receives a "Gold" environmental contribution rating; Tenant C, by participating in demand response (actively reducing unnecessary loads 3 times), receives additional points; Privacy. Protection: Laplace noise with ε=0.5 is added to the energy consumption data of each tenant before transmission. After data aggregation, it is impossible to infer the electricity consumption behavior of a single tenant, thus protecting sensitive commercial information; Dynamic scenario weight: The weather forecast for a certain day is sunny turning cloudy. The system updates the scenario weight of the prediction error every 6 hours. At 8:00 am, strong convective weather actually occurs (with 85% similarity to the 37th extreme scenario in the historical database). The system temporarily increases the weight of this scenario to twice the normal weight. The robust optimization engine improves the safety boundary accordingly, and the energy storage is charged to 95% SOC in advance, successfully coping with the drastic fluctuations in photovoltaic output in the following 2 hours.

[0044] Implementation Scenario 4: Performance monitoring and feedback optimization of long-term evolution.

[0045] After 18 months of continuous operation, the performance monitoring module found that the energy storage battery health index (SOH) dropped from 100% to 92%, and the life prediction model, based on the Wiener process, predicted a remaining cycle life of 800 cycles (an early warning was generated when the threshold was set to 500 cycles). The system automatically adjusted the charging and discharging strategy, narrowing the maximum depth of discharge from 90% to 80%, and suggested increasing ventilation and heat dissipation to extend the life by 10%. Due to the addition of electric vehicle charging piles, the building load increased daily electricity consumption from 2200kWh to 2800kWh. The online learning algorithm (Q-learning) retrained the reinforcement learning strategy, and two weeks later the system found a new optimal strategy: increasing the energy storage capacity reservation ratio from 15% to 25%, and directing charging piles to charge during peak photovoltaic periods, which actually reduced grid interaction costs by 5%. The prediction accuracy index was continuously monitored: the root mean square error of load prediction decreased from the initial 8% to 5%, the photovoltaic prediction error decreased from 12% to 7%, and the error joint distribution model refitted the Gaussian mixture model parameters every 30 days to achieve self-evolution.

[0046] The above implementation scenarios fully demonstrate the specific workflow of this system in the renovation of old buildings, emergency response to extreme weather, multi-entity coordination, and long-term performance optimization, and verify its effectiveness in solving the technical problems of insufficient compatibility of power distribution systems in the renovation of old buildings and weak robustness of prediction errors.

[0047] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A public building grid-connected power distribution system with light storage synergy, characterized in that, include: The prediction error estimation and probability modeling module is used to collect historical data on load prediction error and photovoltaic power generation prediction error in real time, establish a joint probability distribution model of error at multiple time scales, and generate a set of prediction error scenarios for robust optimization. The old building compatibility assessment and progressive upgrade module includes an automated assessment unit, a renovation feasibility report generation unit, a modular hardware platform, and a progressive control algorithm library. It is used to assess the renovation constraints of the target building, generate phased implementation plans, and support phased upgrades from basic monitoring to intelligent optimization. A multi-protocol adaptive interface layer supports automatic identification and conversion of multiple communication protocols, adaptive matching of multi-level voltages, and unified data format processing, enabling seamless integration of legacy equipment with new systems. A robust optimization engine and real-time control actuator include a two-stage stochastic optimization solver, a sub-stochastic optimization engine, an adaptive risk control unit, and a dynamic safety boundary adjustment mechanism. These are used to generate and execute optimal power allocation strategies with robust prediction error. The output of the prediction error estimation and probabilistic modeling module is connected to the input of the robust optimization engine. The output of the old building compatibility assessment and progressive upgrade module is connected to the configuration parameters of the multi-protocol adaptive interface layer. The output of the multi-protocol adaptive interface layer is connected to the constraints of the robust optimization engine. The output of the robust optimization engine is connected to the control commands of the real-time control actuator.

2. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 1, characterized in that: The probability modeling method in the prediction error estimation and probability modeling module Includes the following steps: S1: Collect the load forecast error sequence and photovoltaic power generation forecast error sequence of the past 30 days, and calculate their mean, variance, skewness and kurtosis respectively; S2: Construct a joint probability distribution model of the error, and use a Gaussian mixture model to fit the multimodal distribution characteristics of the error. The model parameters are estimated by the expectation-maximization algorithm; S3: Generate a set of forecast error scenarios: randomly select N scenario samples from the fitted probability distribution. Each scenario contains the error value every 15 minutes in the next 24 hours; S4: Assign probability weights to each scenario, with the weights proportional to the frequency of that scenario in historical data.

3. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 2, characterized in that: The incremental upgrade control algorithm in the old building compatibility assessment and incremental upgrade module includes the following four stages: Stage 1: Monitoring mode, only collects building load, photovoltaic power generation and energy storage status data, establishes a baseline load model, and does not perform any control actions; Phase 2: Assisted mode, which provides power allocation suggestions based on the baseline model, which are then implemented after manual confirmation. The system records the correspondence between the manual decision and the actual effect. The third stage is the semi-automatic mode, which automatically performs power allocation in normal scenarios where the prediction error is less than 15%, and switches to manual intervention mode in extreme scenarios where the prediction error is greater than 15% or the power grid is abnormal; the fourth stage is the fully automatic mode, which trains intelligent control strategies based on reinforcement learning algorithms to achieve adaptive optimization control in all scenarios, and locks the algorithm parameters after the system performance reaches the preset target.

4. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 3, characterized in that: The multi-protocol adaptive interface layer includes: a protocol identification unit, used to monitor the data frame format, baud rate, and parity bit of the communication port in real time, match it with a preset protocol feature library, and automatically identify the current device's communication protocol type; a data conversion engine, used to convert data frames of different protocols into a unified JSON format, with field mapping relationships stored in a configurable mapping table; a voltage adapter, used to detect the device's rated voltage and switch to the matching voltage level through a solid-state relay array, supporting three standards: 220V±10%, 380V±10%, and 10kV±5%; and a plug-and-play interface, using an RJ45 physical interface, supporting hot-swapping, and completing automatic identification and configuration within 30 seconds after device connection.

5. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 4, characterized in that: The two-stage stochastic optimization algorithm in the robust optimization engine is modeled as follows: Objective function: The optimization objective is to minimize the total cost, which includes the expected cost and risk cost of the first-stage decision. The first-stage decision variables include the energy storage charging and discharging plan and the photovoltaic power output plan. The risk cost is the product of the variance of the second-stage adjustment cost and the risk aversion coefficient. The second-stage adjustment cost includes grid interaction cost, load shedding penalty and equipment loss cost. The risk aversion coefficient is dynamically adjusted according to the system operating status. Constraints include: (i) Equipment capacity constraint: The output power of the energy storage battery is limited between its minimum and maximum allowable power, and the state of charge of the energy storage battery is limited. (ii) Power balance constraint: The sum of the output power of the energy storage battery, the output power of the photovoltaic power generation and the power of the grid interaction is equal to the building load power demand; (iii) Operational safety constraint: The power of the grid interaction does not exceed the preset safety boundary; (iv) Prediction error adaptive safety boundary: The safety boundary of the grid interaction power is adaptively adjusted according to the prediction error random variable to ensure that it does not exceed the limit in extreme scenarios; wherein, the prediction error random variable follows a joint probability distribution, the output power of the energy storage battery is positive when it is discharging and negative when it is charging, and the power of the grid interaction is positive when it is drawing power from the grid and negative when it is feeding power to the grid.

6. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 1, characterized in that: The system also includes a performance monitoring and feedback optimization module, which is used to collect system operation indicators in real time, including power supply reliability indicators, economic indicators, equipment health indicators and prediction accuracy indicators, and establish a multi-objective optimization function based on the above indicators, and continuously optimize the control strategy parameters through online learning algorithms.

7. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 1, characterized in that: The system supports three operating modes: grid-connected optimization mode, off-grid autonomous mode, and emergency response mode, which automatically switch according to the grid status and system requirements.

8. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 7, characterized in that: The off-grid autonomous mode includes the following control logic: Off-grid handover detection: if the grid voltage exceeds the rated range by ±10% for three consecutive sampling cycles, off-grid handover is triggered; Critical load identification: Identify the types of loads and power requirements that must be guaranteed within the building based on historical data; Power allocation priority: The first priority is safety lighting, fire protection systems, and medical equipment; The second priority is office equipment and air conditioning systems; the third priority is non-essential loads; energy storage SOC protection: when the energy storage SOC is below 20%, the third priority loads are gradually reduced; when the energy storage SOC is below 10%, the second priority loads are reduced.

9. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 8, characterized in that: The multi-protocol adaptive interface layer also includes a device health status monitoring unit and a lifespan prediction model. Specifically: the device health status monitoring unit collects real-time operating parameters of the access devices, including operating temperature, vibration amplitude, communication error rate, and response latency, and determines the device health status level based on preset threshold rules; the lifespan prediction model uses a survival analysis algorithm to predict the remaining lifespan of the device based on historical operating data, environmental parameters, and load curves. When the predicted remaining lifespan is lower than a preset threshold, it generates an early warning signal and provides maintenance suggestions; the interface layer dynamically adjusts the communication sampling rate and data verification strength according to the device health status level, and for devices with low health status levels, it adds data redundancy verification and retransmission mechanisms.

10. The photovoltaic-storage synergistic public building grid-connected power distribution system as described in claim 9, characterized in that: The scene probability weights in S4 are dynamically adjustable. The adjustment methods include: weight update cycle: the scene weights are updated every 6 hours based on the latest prediction error data; weight adjustment algorithm: a Bayesian update method is used, which uses real-time prediction error observations as new evidence to update the posterior probability of the scene; weight constraint: the weight adjustment of any scene does not exceed ±50% of the weight of the previous cycle to avoid drastic weight fluctuations; extreme scene enhancement: when the similarity between the current weather conditions and a certain historical extreme scene exceeds 80%, the weight of that scene is temporarily increased to twice the normal weight.

Images