New energy storage intelligent scheduling system based on intelligent management platform

Abstract

The invention discloses a new energy storage intelligent scheduling system based on an intelligent management platform, which relates to the technical field of energy monitoring and comprises a data acquisition module, a prediction analysis module, an optimization scheduling module and a strategy execution module. According to the invention, through fusion of precise prediction of the LSTM neural network and the XGBoost model and rolling optimization of the scheduling strategy, intelligent management of the new energy storage system is realized, the technical bottleneck of a traditional scheduling system is broken through, and through multi-dimensional data fusion and dynamic optimization, the accuracy of energy prediction and the adaptability of the scheduling strategy are significantly improved, and the system has good application prospects. The method can accurately capture the periodic characteristics of the photovoltaic output fluctuation and the school load, provides a scientific basis for the charging and discharging decision of an energy storage system, and fundamentally solves the problems of prediction lag and extensive scheduling in a traditional scheme.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy monitoring, in particular to a new energy storage intelligent scheduling system based on a smart management platform. BACKGROUND

[0002] The development of the smart management platform aims to solve the problems of resource dispersion, information silos and inefficient management in environments such as parks and schools. Relying on automation and information technology, it realizes integrated management and control of multiple modules such as fire safety, security, lighting and broadcasting, and integrates environmental and energy monitoring functions. This development background is derived from technological progress, market demand and policy promotion, and seeks innovation in the face of industry pain points. Future trends will develop towards intelligence and automation, focus on cross-domain integration, emphasize green sustainability, and provide customized services to meet market demand and technological innovation. Under the background of deepening the "double carbon" goal, the large-scale application of new energy storage systems in public building scenarios faces the dual challenges of insufficient prediction accuracy and extensive scheduling strategies.

[0003] Currently, traditional energy storage scheduling systems mostly use rule-based control modes, such as using a fixed SOC threshold (e.g. 30%) as the trigger condition for charging and discharging. This approach is difficult to adapt to the intermittent fluctuations of photovoltaic output and the periodic characteristics of school loads. The operation data of a pilot school shows that the traditional BP neural network prediction model it uses has a photovoltaic power prediction error rate of 10.2% under cloudy weather, resulting in 42% of the load still relying on the grid for power supply during peak periods (electricity price 1.2 yuan / kWh), which increases the cost by 28% compared to the theoretical optimal scheduling scheme. At the same time, existing systems generally lack quantitative modeling of device physical constraints, such as not considering the 3% annual power decay characteristic of photovoltaic components and the approximately 0.03% capacity loss of lithium iron phosphate batteries per deep charge and discharge, which causes the available capacity of the energy storage system to decrease to 82% of the rated value after 3 years of actual operation, increasing the equipment operation and maintenance cost by 45% compared to the initial period.

[0004] In view of the problems in the related art, no effective solutions have been proposed so far. SUMMARY

[0005] In view of the problems in the related art, the present application proposes a new energy storage intelligent scheduling system based on a smart management platform to overcome the above technical problems existing in the prior art.

[0006] The technical solution of the present application is as follows:

[0007] A new energy storage intelligent scheduling system based on a smart management platform, comprising: a data acquisition module, a prediction analysis module, an optimization scheduling module and a strategy execution module, wherein;

[0008] The data acquisition module is configured to acquire real-time power generation data of the photovoltaic panel, state data of the energy storage system, and load data of the school power consumption facility through the deployment of the sensor node, and transmit the data to the prediction analysis module through a communication network.

[0009] The prediction analysis module is configured to perform short-term prediction on photovoltaic power by using an LSTM neural network model, and predict power consumption load by using an XGBoost model.

[0010] The optimization scheduling module is configured to construct an energy storage scheduling optimization model with the minimum comprehensive power consumption cost as the target based on a model predictive control algorithm, and generate a charging and discharging strategy of the energy storage system in combination with constraint conditions of the energy storage system.

[0011] The strategy execution module is configured to send the generated charging and discharging strategy to the energy storage system, and control charging and discharging actions of the energy storage system.

[0012] Further, the photovoltaic panel power generation data includes output power, irradiance, and temperature of the photovoltaic panel, the energy storage system state data includes SOC, charging and discharging power, and battery temperature of the energy storage system, and the school power consumption facility load data includes real-time power of each power consumption area.

[0013] Further, the LSTM neural network model adopts a two-layer bidirectional LSTM network structure, each layer has 128 neurons, the input layer is meteorological data and historical 7-day power data, the output layer is a future 1-hour power prediction value, and the time step is 15 minutes.

[0014] Further, the meteorological data includes irradiance, temperature, humidity, and wind speed.

[0015] Further, the objective function of the energy storage scheduling optimization model is represented as:

[0016]

[0017] wherein, C grid (t) is municipal power at time t, P grid (t) is power grid purchase power at time t, C es (t) is depreciation cost of charging and discharging of the energy storage system, ΔE(t) is charging and discharging amount of the energy storage system at time t, R sell (t) is surplus power grid price at time t, P sell (t) is power grid power at time t, t year is service life of the photovoltaic panel.

[0018] Further, the constraint conditions of the energy storage system include:

[0019] Energy storage capacity constraint: SOC min≤ SOC(t) ≤ SOC max ; wherein, SOC min is the minimum SOC value of the energy storage system, SOC max is the maximum SOC value of the energy storage system;

[0020] Charge-discharge power constraint: wherein, P es,max is the minimum charge-discharge power of the energy storage system, P es,max is the maximum charge-discharge power of the energy storage system, and Δt is the scheduling period;

[0021] Energy balance constraint: E(t) = E(t-1) + ΔE(t)·Δt; wherein, E(t) is the energy of the energy storage system at t time;

[0022] Photovoltaic attenuation constraint: P pv (t) ≤ P pv rated × (1-0.003×t year ); wherein, P pv (t) is the actual power of the photovoltaic panel at t time, P pv rated is the rated power of the photovoltaic panel, and t year is the service life of the photovoltaic panel;

[0023] Battery cycle life constraint:

[0024] wherein, N cycle is the cycle number of the battery, and E es,max is the maximum capacity of the energy storage system.

[0025] Further, the model predictive control algorithm comprises: adopting a rolling optimization mechanism, using an LSTM neural network model and an XGBoost model to obtain photovoltaic power and load prediction data for the next 4 hours at t time, solving an energy storage scheduling optimization model, generating a charge-discharge strategy for the t to t+16 period, executing the scheduling instruction at t time, and correcting the energy storage scheduling optimization model based on the latest measured data at t+15 minutes.

[0026] Further, the strategy execution module further comprises: when the SOC of the energy storage system is lower than a preset threshold, automatically switching to a power grid priority power supply mode; and when photovoltaic power sudden drop exceeds a preset threshold, starting an emergency discharge plan.

[0027] Advantages of the present application:

[0028] 1、The application realizes intelligent management of new energy energy storage systems through precise prediction of the fusion of LSTM neural networks and XGBoost models and rolling optimization scheduling strategy, breaks through the technical bottleneck of traditional scheduling systems, significantly improves the accuracy of energy prediction and the adaptability of scheduling strategy through multi-dimensional data fusion and dynamic optimization, can accurately capture the periodic characteristics of photovoltaic output fluctuation and school load, provides a scientific basis for the charging and discharging decision of the energy storage system, and fundamentally solves the problems of prediction lag and extensive scheduling in traditional schemes.

[0029] 2、The application builds a multi-objective optimization model of minimizing comprehensive electricity cost, fully considers the physical constraints and life cycle cost of the energy storage system, realizes low-cost and efficient scheduling while effectively ensuring the reliability of equipment operation, continuously corrects the prediction error through the rolling optimization scheduling strategy, ensures that the energy storage SOC is always maintained in a reasonable range, improves the utilization rate of photovoltaic energy, prolongs the service life of the energy storage equipment, realizes the synergistic optimization of energy efficiency, cost control and equipment protection, and provides an innovative and practical technical solution for new energy energy storage application in public building scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0031] Figure 1 It is a principle block diagram of a new energy energy storage intelligent scheduling system based on a wisdom management platform according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art belong to the scope of protection of the present application.

[0033] According to an embodiment of the present application, a new energy energy storage intelligent scheduling system based on a wisdom management platform is provided.

[0034] As shown in Figure 1 , the new energy energy storage intelligent scheduling system based on the wisdom management platform according to the embodiment of the present application comprises:

[0035] The data acquisition module 1 is used for collecting real-time power generation data of the photovoltaic panel, state data of the energy storage system and load data of the school power utilization facility through the deployment of the sensor node, and transmitting to the prediction analysis module 2 through a communication network;

[0036] The collected photovoltaic panel power generation data includes the output power, irradiance and temperature of the photovoltaic panel, the collected energy storage system state data includes the SOC, charging and discharging power and battery temperature of the energy storage system, and the collected school power utilization facility load data includes the real-time power of each power utilization area.

[0037] Specifically, in the application, single-crystal silicon photovoltaic panels are deployed on the roof and open space of the school, and the photovoltaic panels are arranged at an interval of 1000 m 2 The playground can be installed with a standard configuration of 200kW photovoltaic array, matched with MPPT (maximum power point tracking) controller, and the conversion efficiency is ≥98.5%. The photovoltaic panel adopts 540Wp single block component, and the arrangement inclination is adjusted according to the local latitude (such as north latitude 30°) ±5°, to ensure that the annual power generation hours are ≥1100h. At the same time, the iron lithium phosphate energy storage power station is deployed as an energy storage system, and the capacity is configured according to 40% of the daily average power consumption (such as 1000kWh per day, and 500kWh energy storage battery is selected), matched with a bidirectional converter (PCS) with a rated power of 50kW, and the charging and discharging efficiency is ≥95%. The battery management system (BMS) monitors the single cell voltage and temperature in real time, the sampling frequency is 10Hz, and the SOC estimation error is ≤2%.

[0038] In addition, the school power utilization facility is configured, the power utilization equipment of the teaching building, the canteen and the like is intelligently transformed, and the Hall current sensor (accuracy 0.5 level) and the intelligent electric meter (supporting Modbus-RTU protocol) are respectively installed, to realize the power utilization data acquisition at an interval of 15 minutes; and the air conditioner and the lighting system are connected to the DDC controller, to support remote start and stop and power regulation, the communication network adopts LoRa wireless transmission, and the data delay is ≤50ms.

[0039] The prediction analysis module 2 is used for short-term prediction of photovoltaic power by using an LSTM neural network model, and prediction of power utilization load by using an XGBoost model.

[0040] The LSTM neural network model is built for photovoltaic power prediction, including adopting a 2-layer bidirectional LSTM network structure, 128 neurons per layer, and the specific parameters are as follows:

[0041] Input layer: meteorological data (irradiance, temperature, humidity, wind speed) + historical 7-day power data (15-minute resolution);

[0042] Output layer: future 1-hour power prediction value, time step 15 minutes;

[0043] Training parameters: Adam optimizer, learning rate 0.001, batch size 32, 500 iterations, loss function mean squared error (MSE).

[0044] where the mathematical model is denoted as:

[0045] Let the input sequence be X t = [I t , T t , H t , W t, , P pv,t-1 ,..., P pv,t-288 ], where I t is irradiance, T t is temperature, and the time span is 7 days x 96 points / day = 288 points, then the LSTM forward propagation process is represented as:

[0046] i t = σ(W xi X t + W hi H t-1 + W ci C t-1 + b i );

[0047] f t = σ(W xf X t + W hf H t-1 + W cf C t-1 + b f );

[0048] C t = f t C t-1 + i t tanh(W xc X t + W hc H t-1 + b c );

[0049] o t = σ(W xo X t + W ho H t-1 + W co C t + b o );

[0050] H t = o t tanh(C t );

[0051]

[0052] where σ is the Sigmoid activation function, W is the weight matrix, b is the bias vector, i t , f t , o t are the input gate, forget gate, output gate, respectively, C t is the cell state, and H t is the hidden state.

[0053] where the XGBoost model is built for electricity load forecasting, including: building an integrated learning model containing time characteristics and scene characteristics, specific parameters:

[0054] Time characteristics: weekdays / weekends (one-hot encoding), 24-hour period (0-23), season (spring / summer / autumn / winter);

[0055] Scene characteristics: school calendar (school day / holiday, data from school OA system), large event day (obtained through API);

[0056] Model parameters: tree depth 6, learning rate 0.1, iteration number 100, regularization parameter λ=0.1, γ=0.5.

[0057] where the mathematical model is represented as:

[0058]

[0059] where the loss function l adopts Huber loss, which is suitable for processing outliers;

[0060] The regularization term Ω(f k ) is represented as:

[0061]

[0062] where T is the number of leaf nodes, and w j is the weight of the jth leaf node.

[0063] The optimization scheduling module 3, based on the model predictive control algorithm, builds an energy storage scheduling optimization model with the minimum comprehensive electricity cost as the target, and generates the charging and discharging strategy of the energy storage system combined with the constraint conditions of the energy storage system;

[0064] where the energy storage scheduling optimization model is built, including calibrating the minimum comprehensive electricity cost as the core target, while satisfying the energy storage SOC constraint, building the objective function, which is represented as:

[0065]

[0066] where Cgrid (t) is the municipal electricity at time t, taking the value of 1.2 yuan / kWh at the peak and 0.5 yuan / kWh at the valley, P grid (t) is the grid purchase electricity at time t, C es (t) is the depreciation cost of the energy storage system charging and discharging, taking the value of 0.15 yuan / kWh, ΔE(t) is the charging and discharging capacity of the energy storage system at time t, charging is negative and discharging is positive, R sell (t) is the excess electricity grid price at time t, taking the value of 0.4 yuan / kWh, P sell (t) is the grid electricity at time t, t year is the service life of photovoltaic.

[0067] Among them, the constraint condition is represented as:

[0068] Energy storage capacity constraint: SOC min ≤ SOC(t) ≤ SOC max ; wherein, SOC min is the minimum SOC value of the energy storage system, SOC max is the maximum SOC value of the energy storage system;

[0069] Charging and discharging power constraint: wherein, P es,max is the minimum charging and discharging power of the energy storage system, P es,max is the maximum charging and discharging power of the energy storage system, and Δt is the scheduling period;

[0070] Energy balance constraint: E(t) = E(t-1) + ΔE(t)·Δt; wherein, E(t) is the energy of the energy storage system at time t;

[0071] Photovoltaic attenuation constraint: P pv (t) ≤ P pv rated×(1-0.003×t year ); wherein, P pv (t) is the actual power of the photovoltaic panel at time t, P pv rated is the rated power of the photovoltaic panel, t year is the service life of the photovoltaic panel;

[0072] Battery cycle life constraint:

[0073] wherein, N cycle is the number of cycles of the battery, E es,max is the maximum capacity of the energy storage system.

[0074] Among them, the model predictive control algorithm adopts a rolling optimization mechanism, and the specific steps are as follows:

[0075] At time t, the future 4 hours (16 scheduling periods) of photovoltaic power and load prediction data are obtained by using the LSTM neural network model and the XGBoost model;

[0076] The energy storage scheduling optimization model is solved to generate the charging and discharging strategy from t to t+16 periods;

[0077] The scheduling instruction at time t is executed, and P grid (t) and ΔE(t) are sent to the energy storage system;

[0078] At t+15 minutes, the energy storage scheduling optimization model is corrected based on the latest measured data, and the above steps are repeated to suppress error accumulation.

[0079] Specifically, the scheduling period of the model predictive control algorithm is 15 minutes, and the charging and discharging strategy for the next 4 hours is optimized, and after each scheduling period is completed, the energy storage scheduling optimization model is updated according to the latest measured data, and the subsequent charging and discharging strategy is regenerated.

[0080] The strategy execution module 4 is used to send the generated charging and discharging strategy to the energy storage system to control the charging and discharging action of the energy storage system.

[0081] When the SOC of the energy storage system is lower than a preset threshold, it is automatically switched to a power grid priority power supply mode; when the photovoltaic power drops by more than a preset threshold, an emergency discharge plan is started.

[0082] As described above, by means of the above technical scheme of the present application, the following effects can be achieved:

[0083] 1. The present application realizes intelligent management of new energy storage systems by fusing the precise prediction of LSTM neural network and XGBoost model and rolling optimization of scheduling strategy, breaks through the technical bottleneck of traditional scheduling systems, significantly improves the accuracy of energy prediction and the adaptability of scheduling strategy through multi-dimensional data fusion and dynamic optimization, accurately captures the periodic characteristics of photovoltaic output fluctuation and school load, and provides a scientific basis for the charging and discharging decision of the energy storage system, fundamentally solving the problems of prediction lag and extensive scheduling in traditional schemes.

[0084] 2. The present application builds a multi-objective optimization model that minimizes comprehensive electricity cost, fully considers the physical constraints and life cycle cost of the energy storage system, realizes low-cost and efficient scheduling while effectively ensuring the reliability of equipment operation, continuously corrects prediction errors through rolling optimization of scheduling strategy, ensures that the energy storage SOC is always maintained in a reasonable range, improves the utilization rate of photovoltaic energy, prolongs the service life of the energy storage equipment, realizes the coordinated optimization of energy efficiency, cost control and equipment protection, and provides an innovative and practical technical solution for new energy storage applications in public building scenarios.

[0085] The foregoing is considered as illustrative only of the principles of the disclosure. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the disclosure to the exact construction and procedure described. Accordingly, all such variations are intended to be included within the scope of the present disclosure as defined in the claims below. The disclosure is also not limited to the details given in the description above and the drawings, but can vary within the scope of the claims.

[0086] It is to be understood that the specific devices and methods described above are merely illustrative of the principles of the disclosure and that numerous modifications and changes can be made without departing from the scope of the disclosure. The scope of the disclosure is limited only by the claims below.

Claims

1. A new energy storage intelligent dispatching system based on a smart management platform, characterized in that, include: The system comprises a data acquisition module (1), a predictive analysis module (2), an optimization scheduling module (3), and a strategy execution module (4), wherein; The data acquisition module (1) is used to collect real-time power generation data of photovoltaic panels, status data of energy storage system and load data of school power facilities by deploying sensor nodes, and transmit them to the predictive analysis module (2) through the communication network. The prediction and analysis module (2) is used to make short-term predictions of photovoltaic power using the LSTM neural network model and to predict electricity load using the XGBoost model. The optimization scheduling module (3) constructs an energy storage scheduling optimization model with the goal of minimizing the comprehensive electricity cost based on the model predictive control algorithm, and generates the charging and discharging strategy of the energy storage system in combination with the constraints of the energy storage system. The strategy execution module (4) is used to send the generated charging and discharging strategy to the energy storage system and control the charging and discharging actions of the energy storage system.

2. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 1, characterized in that, The photovoltaic power generation data includes the output power, irradiance, and temperature of the photovoltaic panel; the energy storage system status data includes the SOC, charging and discharging power, and battery temperature of the energy storage system; and the school's power facility load data includes the real-time power of each power consumption area.

3. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 1, characterized in that, The LSTM neural network model adopts a two-layer bidirectional LSTM network structure with 128 neurons in each layer. The input layer contains meteorological data and historical 7-day power data, and the output layer contains the power prediction value for the next hour. The time step is 15 minutes.

4. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 3, characterized in that, The meteorological data includes irradiance, temperature, humidity, and wind speed.

5. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 1, characterized in that, The objective function of the energy storage scheduling optimization model is expressed as: Among them, C grid (t) represents the municipal electricity at time t, P grid (t) represents the electricity purchased by the power grid at time t, C es (t) represents the depreciation cost of the energy storage system during charging and discharging, ΔE(t) represents the charging and discharging amount of the energy storage system at time t, and R sell (t) represents the surplus electricity price at time t, P sell (t) represents the electricity consumed during internet use at time t. year This refers to the lifespan of the photovoltaic system.

6. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 5, characterized in that, The constraints of the energy storage system include: Energy storage capacity constraints: SOC min ≤SOC(t)≤SOC max Among them, SOC min The minimum SOC value for the energy storage system. max This represents the maximum SOC value of the energy storage system. Charge and discharge power constraints: Among them, P es,max P is the minimum charge / discharge power of the energy storage system. es,max Δt represents the maximum charging and discharging power of the energy storage system, and Δt represents the scheduling period. Energy balance constraint: E(t) = E(t-1) + ΔE(t)·Δt; where E(t) is the energy of the energy storage system at time t; Photovoltaic degradation constraint: P pv (t)≤P pv ,rated×(1-0.003×t year ); where P pv (t) represents the actual power of the photovoltaic panel at time t, P pv ,rated is the rated power of the photovoltaic panel, t year The service life of photovoltaic panels; Battery cycle life constraints: Where, N cycle E represents the number of battery cycles. es,max This represents the maximum capacity of the energy storage system.

7. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 1, characterized in that, The model predictive control algorithm includes: using a rolling optimization mechanism, at time t, using an LSTM neural network model and an XGBoost model to obtain photovoltaic power and load prediction data for the next 4 hours, solving the energy storage scheduling optimization model, generating a charging and discharging strategy for cycles from t to t+16, executing the scheduling command at time t, and at t+15 minutes, correcting the energy storage scheduling optimization model based on the latest measured data.

8. The intelligent dispatching system for new energy storage based on a smart management platform according to claim 1, characterized in that, The strategy execution module (4) further includes: automatically switching to grid priority power supply mode when the SOC of the energy storage system is lower than a preset threshold; and activating an emergency discharge plan when the photovoltaic power drops sharply beyond a preset threshold.

Images