Photovoltaic energy storage motor cooperative scheduling optimization system based on multi-modal data

Abstract

The invention relates to the technical field of energy management and intelligent scheduling, provides a photovoltaic energy storage motor collaborative scheduling optimization system based on multi-modal data, and constructs a collaborative scheduling system integrating data intelligent processing, high-precision prediction and multi-objective optimization by fusing photovoltaic module, energy storage equipment, motor and power grid side multi-modal data. Data quality is improved through Z-score cleaning and Kalman filtering, accurate prediction of photovoltaic power, load demand and energy storage aging is realized by using a long-short-term memory neural network and an Arrhenius equation, and a multi-target collaborative scheduling strategy including economic, reliable and environmental benefits is constructed based on a particle swarm optimization algorithm. Dynamically adjusting photovoltaic output distribution, energy storage charging and discharging and a motor speed regulation ratio; systematic innovation of data value deep mining, multi-target dynamic balance, equipment full-life-cycle management and renewable energy source efficient consumption is realized, and the intelligent level and comprehensive operation efficiency of the micro-grid are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy management and intelligent scheduling, and specifically to a photovoltaic energy storage motor collaborative scheduling optimization system based on multimodal data. Background Art

[0002] With the widespread adoption of renewable energy, especially the increasing share of photovoltaic power generation systems in energy supply, their intermittent and unstable nature poses challenges to the stable operation of energy systems. Furthermore, the aging performance of energy storage devices and the dynamic demand changes of load devices such as motors further exacerbate the complexity of system operation. Existing energy management systems lack sufficient data collection, processing, and scheduling strategies, and are unable to accurately predict photovoltaic power generation, load demand, and the aging characteristics of energy storage equipment. This results in high system operating costs, poor reliability, and low environmental benefits.

[0003] Therefore, a photovoltaic energy storage and motor coordinated scheduling optimization system based on multimodal data was proposed. High-precision sensors are used to collect multi-dimensional data from photovoltaic modules, energy storage devices, motors, and the grid in real time to construct a comprehensive raw data set. Advanced data processing and analysis techniques are used to clean and fuse the collected multimodal data to obtain an accurate estimate of the system status. On this basis, deep learning algorithms are used to accurately predict future photovoltaic power generation, load demand, and aging characteristics of energy storage equipment. In combination with multi-objective optimization algorithms, an economical, reliable, and environmentally friendly scheduling strategy is formulated. The system also has time-sharing and partitioned scheduling and multi-device coordinated control functions. It can flexibly adjust operating strategies according to different time periods and equipment status, effectively addressing problems such as sudden changes in photovoltaic power and aging of energy storage equipment, significantly improving the overall performance and economic benefits of the system, and meeting the needs of modern energy systems for intelligent and efficient operation. Summary of the Invention

[0004] Technical problems solved

[0005] In response to the deficiencies of the prior art, the present invention provides a photovoltaic energy storage motor collaborative scheduling optimization system based on multimodal data.

[0006] Technical Solution

[0007] To achieve the above-mentioned solution, the present invention provides the following technical solution: a photovoltaic energy storage motor collaborative scheduling optimization system based on multimodal data, including a data acquisition module, a data processing and analysis module, a prediction module, a scheduling strategy optimization module and a system collaborative scheduling mechanism.

[0008] The data acquisition module is used to collect photovoltaic module data, energy storage device data, motor data and grid-side data in real time through multiple sensors to construct the original data set; it includes a photovoltaic module data acquisition unit, an energy storage device data acquisition unit, a motor data acquisition unit and a grid-side data acquisition unit.

[0009] The data processing and analysis module is used to clean and fuse the collected multimodal data to obtain an accurate estimate of the system state; it includes a data cleaning unit and a data fusion unit.

[0010] The prediction module is used to predict future photovoltaic power generation power, load demand and energy storage equipment aging characteristics based on the processed data; it includes a photovoltaic power generation power prediction unit, a load demand prediction unit and an energy storage equipment aging characteristics prediction unit.

[0011] The scheduling strategy optimization module is used to construct a multi-objective function and optimize the scheduling strategy using an intelligent optimization algorithm; it includes a multi-objective function construction unit and an intelligent optimization algorithm unit.

[0012] The system collaborative scheduling mechanism is used to execute corresponding time-sharing and partition scheduling strategies and multi-device collaborative control operations according to different time periods and device states; it includes a time-sharing and partition scheduling strategy unit and a multi-device collaborative control unit.

[0013] Preferably, the photovoltaic module data acquisition unit adopts a high-precision power sensor, a light intensity sensor and a temperature sensor to collect the output power, light intensity and temperature data of the photovoltaic module in real time at a frequency of 15 minutes / time; the energy storage device data acquisition unit obtains the charge state, charge and discharge current and voltage data of the energy storage device through the BMS, with a sampling frequency of 5 minutes / time; the motor data acquisition unit uses a vibration sensor, a current sensor and a speed sensor to collect the speed, load and current data of the motor, with a sampling frequency of 1 minute / time; the grid-side data acquisition unit obtains the voltage, frequency and real-time electricity price data of the grid side from the smart meter, with a sampling frequency of 15 minutes / time.

[0014] Preferably, the data cleaning unit adopts the Z-score method to eliminate abnormal data, and obtains the Z-score by calculating the difference between the data value and the mean and dividing it by the standard deviation. The threshold is pre-set to 3 or -3. When the Z-score of the data exceeds the threshold range, the data is determined to be abnormal data and is eliminated; the data fusion unit adopts the Kalman filter algorithm to realize data fusion, estimates the system state through the prediction step and the update step, and reduces the data error.

[0015] Preferably, the photovoltaic power generation prediction unit adopts a three-layer LSTM network combined with a fully connected layer to output the predicted value. The input features include historical photovoltaic power, light intensity, temperature, humidity, and wind speed. The output is a power sequence for the next 1 to 24 hours with a time resolution of 15 minutes. The load demand prediction unit also adopts an LSTM neural network. The input features include historical motor load, production plan, and time factor. The output is a load sequence for the next 1 to 12 hours with a time resolution of 15 minutes. The energy storage device aging characteristic prediction unit is based on the Arrhenius equation coupling model and combines the effects of temperature and charge and discharge depth on battery aging to predict and establish a remaining service life prediction model.

[0016] Preferably, the multi-objective function construction unit constructs an economic objective function, a reliability objective function and an environmental benefit objective function, and forms a comprehensive objective function through balancing weight coefficients; the intelligent optimization algorithm unit adopts a particle swarm optimization algorithm, and the particle encoding includes three decision variables: photovoltaic output distribution coefficient, energy storage charging and discharging power, and motor speed regulation ratio. The algorithm process includes initializing the particle swarm, calculating fitness, updating individual optimality and global optimality, speed and position updating, and constraint processing.

[0017] Preferably, the time-sharing and partitioning scheduling strategy unit executes different scheduling strategies for photovoltaic peak periods, photovoltaic valley periods, and transition periods to optimize system operation; the multi-device collaborative control unit is used to handle photovoltaic power mutations and energy storage equipment aging to ensure that the system can still operate reliably in the face of emergencies and extend the service life of the energy storage equipment.

[0018] Preferably, during the photovoltaic peak period, photovoltaic power is preferentially allocated to the motor, and the remaining photovoltaic power is used to charge the energy storage device. When the SOC of the energy storage device reaches a fully charged state, the motor speed regulation ratio is intelligently adjusted to increase the motor load and maximize the absorption of photovoltaic power. During the photovoltaic off-peak period, intelligent decisions are made on whether to discharge the energy storage device or purchase electricity from the power grid based on the load demand forecast results and real-time electricity price information. During the low electricity price period, low-priced electricity resources are used to charge the energy storage device. During the transition period, the SOC of the energy storage device and the speed regulation ratio of the motor are adjusted in advance based on the forecast data to smooth out power fluctuations and ensure a smooth transition of the system.

[0019] Preferably, the photovoltaic power mutation processing triggers the energy storage device's rapid discharge compensation mechanism when the photovoltaic power drops by more than 20%, and dynamically adjusts the motor speed regulation ratio at the same time, reduces the motor speed by 10%, and reduces power demand; the energy storage device aging processing automatically adjusts the scheduling strategy when it is monitored that the remaining service life of the energy storage device is less than 1 year, limits the DOD to within 50%, reduces the battery aging rate, and extends the service life.

[0020] Beneficial effects

[0021] Compared with the existing technology, the present invention provides a photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data, which has the following beneficial effects:

[0022] 1. This solution integrates multimodal data from photovoltaic modules, energy storage equipment, motors, and the grid to build a collaborative scheduling system that integrates intelligent data processing, high-precision prediction, and multi-objective optimization. Z-score cleaning and Kalman filtering are used to improve data quality. Long-short-term memory neural networks and the Arrhenius equation are used to accurately predict photovoltaic power, load demand, and energy storage aging. A multi-objective collaborative scheduling strategy with economic, reliable, and environmental benefits is constructed based on the particle swarm optimization algorithm to dynamically adjust photovoltaic output distribution, energy storage charging and discharging, and motor speed regulation ratios.

[0023] 2. This solution has overcome technical bottlenecks such as inefficient use of multi-source data, insufficient adaptability of prediction models, single scheduling strategies, and lack of equipment coordination. It has achieved systematic innovation in deep mining of data value, dynamic balance of multiple objectives, full life cycle management of equipment, and efficient absorption of renewable energy, significantly improving the intelligence level and overall operational efficiency of microgrids. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Schematic diagram of the system framework of the present invention;

[0025] Figure 2 Schematic diagram of the system operation flow of the present invention. DETAILED DESCRIPTION

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0027] See also Figures 1 and 2 The present invention proposes a photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data, which includes the following contents:

[0028] 1. Data acquisition module

[0029] (1) Multimodal data acquisition system

[0030] High-precision sensors are used to collect four types of key data in real time to construct the original data set:

[0031] Photovoltaic module data collection: Use high-precision power sensors, light intensity sensors and temperature sensors to collect relevant data of photovoltaic modules in real time. The collected data includes output power (unit: kW), light intensity (unit: W / m 2 ) and temperature (in °C). These sensors sample data every 15 minutes, building a detailed raw data set for PV modules. By collecting and analyzing this data, accurate predictions of PV power and a comprehensive assessment of PV module performance are possible. This not only helps understand the power generation capacity of PV modules under different environmental conditions but also enables timely identification of potential performance issues, providing a solid data foundation for subsequent scheduling strategy development.

[0032] Energy storage device data collection: The data collection of energy storage devices is completed through BMS (battery management system). BMS can accurately obtain the state of charge (SC, unit: %), charge and discharge current (unit: A) and voltage (unit: V) of the energy storage device. In order to capture the changes in the state of the energy storage device in a timely manner, the sampling frequency set in the present invention is 5 minutes / time. These data are crucial for accurately evaluating the current state of the energy storage device and predicting its aging trend. By monitoring the SC, the charging and discharging plan of the energy storage device can be reasonably arranged; tracking the charge and discharge current and voltage will help to promptly detect possible abnormal conditions of the battery, such as overcharging, over-discharging, etc., and then take corresponding measures to ensure the safe and stable operation of the energy storage device and extend its service life.

[0033] Motor data collection: For the motor part, the present invention uses vibration sensors, current sensors and speed sensors to collect data. The collected data covers speed (unit: r / min), load (unit: %) and current (unit: A), and the sampling frequency is set to 1 minute / time. Through real-time monitoring of these data, on the one hand, accurate prediction of motor load demand can be achieved, providing strong support for the power balance of the system; on the other hand, the operating status of the motor can be effectively monitored, and potential fault hazards of the motor, such as abnormal vibration, overload, etc., can be discovered in time, thereby ensuring the reliable operation of the motor and avoiding adverse consequences such as system power outages due to motor failures.

[0034] Grid-side data collection: Grid-side data is collected from smart meters, including voltage (V), frequency (Hz), and real-time electricity prices (yuan / kWh), with a sampling frequency of 15 minutes. This data plays a key role in the system's interaction with the grid. Monitoring grid voltage and frequency ensures good compatibility and stability with the grid during operation. Real-time electricity price information provides a basis for developing economically sound grid interaction strategies, enabling the system to meet its own energy needs while minimizing electricity purchase costs.

[0035] (2) Data transmission and storage

[0036] The collected multimodal data is transmitted to the edge computing server via industrial bus protocols such as Mdbus and CAN. On the server side, this raw data undergoes preliminary preprocessing, including format unification and data integrity verification, to ensure data quality and usability. The preprocessed data is stored in a professional time series database (such as InfluxDB), forming a complete historical dataset. This historical dataset not only provides a rich data sample for subsequent model training, helping to improve the accuracy and reliability of the predictive model, but also provides strong support for the system's retrospective analysis. By deeply mining and analyzing historical data, potential patterns and problems in the system's operation can be discovered, providing a decision-making basis for continuous system optimization.

[0037] 2. Data processing and analysis module

[0038] (1) Data cleaning: Z-score outlier detection

[0039] The Z-score method is used to eliminate abnormal data. The formula is as follows:

[0040]

[0041] Where X represents the data value, μ represents the mean, and σ represents the standard deviation. In practice, a reasonable threshold is pre-set, typically 3 or -3. If the Z-score exceeds this threshold, the data is considered outlier and removed. This method, by normalizing the data, effectively identifies and removes outliers caused by external factors such as sensor failure and communication interference. The cleaned dataset is more reliable and accurate, laying a solid foundation for subsequent data fusion and analysis, ensuring stable operation of the entire system and accurate decision-making.

[0042] (2) Data fusion: Kalman filter algorithm

[0043] To address the spatiotemporal asynchrony problem of multimodal data, the present invention innovatively uses the Kalman filter algorithm to achieve data fusion to obtain an accurate estimate of the system state. The algorithm is mainly divided into a prediction step and an update step.

[0044] In the prediction step, the system state is estimated based on the state transfer matrix and the control input matrix, and the estimation error covariance matrix is ​​updated at the same time. The specific formula is:

[0045]

[0046] In the update step, the Kalman gain is calculated based on the measurement value and the measurement matrix, and the state estimate and the estimation error covariance matrix are further corrected. The relevant formula is as follows:

[0047]

[0048] P k|k =(IK k H)P k|k-1

[0049] in, is the state estimation value, A is the state transfer matrix, B is the control input matrix, u k is the control quantity, P k|k-1 is the estimation error covariance matrix, Q is the process noise covariance matrix, K k is the Kalman gain, z k is the measurement value, H is the measurement matrix, and R is the measurement noise covariance matrix.

[0050] Through this recursive estimation process, the Kalman filter algorithm can effectively fuse multimodal data from different sensors, reducing data errors by 30% to 50%, thereby more accurately reflecting the actual operating status of the system and providing a reliable basis for subsequent prediction and scheduling strategy formulation.

[0051] 3. Prediction Module

[0052] (1) Photovoltaic power generation prediction: LSTM neural network

[0053] Model Architecture: A three-layer LSTM network (128 neurons per layer) is combined with a fully connected layer to output predicted values. Input features include historical photovoltaic power (previous 24 hours), light intensity, temperature, humidity, and wind speed; the output is a power series for the next 1 to 24 hours (with a time resolution of 15 minutes).

[0054] Gating mechanism: Controls the flow of information through input gate, forget gate, and output gate to solve the vanishing gradient problem of traditional RNN. The core formula is as follows:

[0055] it =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i )

[0056] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f )

[0057] c t =f t c t-1 +i t tanh(W xc x t +W hc h t-1 +b c )

[0058] o t =σ(W xo x t +W ho h t-1 +W co c t +b o )

[0059] h t =o t tanh(c t )

[0060] Among them, i t is the input gate output, f t is the output of the forget gate, c t is the cell state, o t is the output gate output, h t is the hidden state, σ is the sigmoid activation function, W is the weight matrix, and b is the bias term.

[0061] Training Strategy: The Adam optimizer was used, with the root mean square error (RMSE) as the loss function. The training set consisted of historical data from the past year, and the validation set consisted of data from the past three months. Test results showed that the model's prediction errors for sunny, cloudy, and rainy days were less than 5%, 8%, and 12%, respectively, representing an improvement of over 40% over traditional ARIMA models.

[0062] (2) Load demand forecasting: LSTM neural network

[0063] Input features: historical motor load (previous 24 hours), production plan (such as work order quantity, equipment start and stop status), time factor (weekday / weekend, hourly period); output is the load sequence for the next 1 to 12 hours (time resolution 15 minutes).

[0064] Model differences: Compared to the PV forecasting model, the input layer adds a production plan encoding vector, and the number of hidden layer neurons is adjusted to 64 to accommodate the high-frequency fluctuations of load data. Field tests at an industrial park demonstrated a prediction error of less than 7%, meeting real-time scheduling requirements.

[0065] (3) Prediction of aging characteristics of energy storage equipment: Arrhenius equation coupling model

[0066] Core formula: Based on the Arrhenius equation, a model is established to model the effects of temperature and depth of charge and discharge (DOD) on battery aging:

[0067]

[0068] Among them, k is the battery aging rate, k0 is the reference aging rate, E a is the activation energy (taken as 0.6eV), k B is the Boltzmann constant (8.617×10 -5 eV / K), T is the actual temperature (K), and T0 is the reference temperature (taken as 25℃, i.e. 298K).

[0069] Coupling analysis: Combined with the relationship between charge and discharge depth and aging (e.g., for every 10% increase in DOD, the cycle life decreases by 15%), a remaining useful life (RUL) prediction model is established:

[0070]

[0071] Among them, R max is the maximum aging degree, R current is the current aging degree, R per cycle The aging degree of each charge and discharge cycle. By real-time monitoring of temperature and DOD, the performance degradation trend of energy storage equipment can be predicted 6 months in advance with an error rate of less than 10%.

[0072] 4. Scheduling strategy optimization module

[0073] (1) Construction of multi-objective functions

[0074] Economic objective function: Minimize system operating costs, including photovoltaic absorption costs, energy storage charging and discharging costs, and grid power purchase costs:

[0075]

[0076] Among them, c p is the photovoltaic power generation cost coefficient (yuan / kWh), P p (t) is the actual photovoltaic power at time t, c s is the energy storage charging and discharging cost coefficient (yuan / A·V), I c (t), V c (t) is the charging current and voltage, I d (t), V d (t) is the discharge current and voltage, e(t) is the electricity price at time t, P g (t) is the power purchased by the power grid, and T is the dispatch period (taken as 24 hours, the time step is 15 minutes, and there are 96 steps in total).

[0077] Reliability objective function: Maximize system power supply reliability, using power outage probability and voltage deviation as evaluation indicators:

[0078]

[0079] Among them, ω1 and ω2 are weight coefficients (range 0-1), P out (t) is the probability of power outage at time t (determined by redundancy design and fault tree analysis), and ΔV(t) is the voltage deviation:

[0080] is the measured voltage, V rated is the rated voltage (380V).

[0081] Environmental benefit objective function: Minimize system carbon emissions, taking into account the carbon emissions from photovoltaic clean power generation and energy storage charging and discharging processes:

[0082]

[0083] Among them, β p is the photovoltaic environmental benefit coefficient (taken as 0.8kgCO2 / kWh), β c , β d is the carbon emission coefficient of energy storage charging and discharging (0.5kgCO2 / A·V for charging and 0.3kgCO2 / A·V for discharging).

[0084] Comprehensive objective function: balancing multiple objective requirements through weight coefficients:

[0085] F=ω1F1+ω2F2+ω3F3

[0086] Among them, ω1+ω2+ω3=1, and users can adjust the weights according to actual needs (for example, the industrial scenario focuses on economy, ω1=0.6; the grid access scenario focuses on reliability, ω2=0.5).

[0087] (2) Intelligent optimization algorithm: Particle Swarm Optimization (PSO)

[0088] Particle encoding: Each particle represents a scheduling strategy solution, which contains three decision variables:

[0089] Photovoltaic output allocation coefficient α(t): the ratio of photovoltaic power allocated to the motor and energy storage (0≤α(t)≤1).

[0090] Energy storage charging and discharging power P s (t): Positive value indicates charging, negative value indicates discharging (limited to the rated power range of energy storage).

[0091] Motor speed regulation ratio γ(t): Motor speed regulation coefficient (0.8≤γ(t)≤1.2, corresponding to load regulation range 60%~100%).

[0092] Algorithm flow:

[0093] Initialize the particle swarm: randomly generate N particles (take N = 100), and the position and velocity of each particle are randomly distributed in the feasible domain.

[0094] Calculate fitness: Calculate the comprehensive objective function value according to the particle position as fitness evaluation.

[0095] Update individual optimal and global optimal: compare the current fitness of the particle with the historical optimal fitness, update the individual optimal position pbest; compare the fitness of all particles, and update the global optimal position gbest.

[0096] Velocity and position updates:

[0097]

[0098] in, is the d-dimensional velocity of the i-th particle in the k-th iteration, ω is the inertia weight (dynamically adjusted to 0.8-0.9), c1 and c2 are learning factors (both 2), r1 and r2 are random numbers in the range [0,1], is the d-dimensional position of the i-th particle in the k-th iteration.

[0099] Constraint processing: Correct the particle positions that exceed the feasible region (such as forced truncation when α(t) exceeds [0,1]).

[0100] Termination condition: When the number of iterations reaches the set value (taken as 200 times) or the fitness value converges, the global optimal solution is output.

[0101] Algorithm advantages: Compared with genetic algorithms, PSO algorithms have fewer parameters and faster convergence speed. In this system, the optimal solution can be reached after an average of 150 iterations, and the calculation time is less than 5 seconds, meeting real-time scheduling requirements.

[0102] 5. System collaborative scheduling mechanism

[0103] (1) Time-sharing and partitioning scheduling strategy

[0104] Photovoltaic peak hours (9:00-15:00)

[0105] During this period, the photovoltaic panel's power generation reaches its peak. The system prioritizes allocating photovoltaic power to the motor to meet its operating needs. The remaining photovoltaic power is used to charge the energy storage device, thereby reducing the need to purchase electricity from the grid and lowering operating costs. If the energy storage device's SOC (State of Charge) reaches full charge, the system intelligently adjusts the motor speed ratio, appropriately increasing the motor load within the device's permitted performance range. This maximizes the absorption of photovoltaic power, fully utilizes renewable energy, and improves the system's economic and environmental performance.

[0106] Photovoltaic off-peak period (18:00-6:00 the next day)

[0107] During this time, based on the system's load demand forecast and real-time electricity price information, the system will intelligently decide whether to discharge energy storage devices or purchase electricity from the grid. Especially during periods of low electricity prices (e.g., 11:00 PM to 7:00 AM), the system will fully utilize low-priced electricity resources to charge energy storage devices, reducing overall system operating costs and achieving the economic benefits of peak load shifting. This will also help balance the load distribution of the grid and improve its operational stability.

[0108] Transition period (6:00-9:00, 15:00-18:00)

[0109] During these transition periods, the system will proactively adjust the energy storage device's SOC (State of Charge) and the motor's speed ratio based on forecast data. By rationally planning the energy storage device's charging and discharging processes and the motor's operating status, power fluctuations are effectively smoothed, ensuring a smooth transition between different operating periods. This prevents sudden power fluctuations from impacting equipment or negatively impacting the grid, ensuring stable and reliable system operation.

[0110] (2) Multi-device collaborative control

[0111] Photovoltaic power sudden change processing

[0112] When PV power suddenly drops by more than 20% due to external environmental factors (such as cloud cover), the system will quickly respond and trigger the energy storage device's rapid discharge compensation mechanism to compensate for the momentary loss of PV power and maintain the continuity and stability of the system's power supply. At the same time, the system will dynamically adjust the motor speed ratio, such as appropriately reducing the motor speed (for example, by 10%), to reduce the motor's power demand to a certain extent, further alleviating the system's power balance pressure, ensuring that the entire system can continue to operate reliably in the face of emergencies and ensuring uninterrupted production activities.

[0113] Energy storage equipment aging treatment

[0114] When monitoring indicates that the energy storage equipment has reached a certain level of aging and its remaining service life is less than one year, the system will automatically adjust its scheduling strategy. Specific measures include reducing the number of deep charge and discharge cycles of the energy storage equipment, limiting the DOD (depth of charge and discharge) to less than 50%. This reduces the battery's aging rate, extends its service life, and improves the return on investment of the energy storage equipment. This not only helps ensure stable operation of the system throughout the energy storage equipment's lifecycle, but also reduces the high costs associated with frequent replacement of energy storage equipment, improving the overall cost-effectiveness and reliability of the system.

[0115] Finally, it should be noted that the above embodiments are merely examples for the purpose of illustrating the present invention and are not intended to limit the embodiments. Those skilled in the art will readily appreciate that other variations or modifications based on the above description are possible. It is not necessary and impossible to provide an exhaustive list of all embodiments. However, obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. Photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data, characterized by: It includes data acquisition module, data processing and analysis module, prediction module, scheduling strategy optimization module and system collaborative scheduling mechanism; The data acquisition module is used to collect PV module data, energy storage device data, motor data, and grid-side data in real time through multiple sensors to construct an original data set; it includes a PV module data acquisition unit, an energy storage device data acquisition unit, a motor data acquisition unit, and a grid-side data acquisition unit; The data processing and analysis module is used to clean and fuse the collected multimodal data to obtain an accurate estimate of the system status; Including data cleaning unit and data fusion unit; A prediction module is used to predict future photovoltaic power generation, load demand, and energy storage equipment aging characteristics based on processed data; It includes photovoltaic power generation prediction unit, load demand prediction unit and energy storage equipment aging characteristics prediction unit; Scheduling strategy optimization module, used to construct multi-objective functions and optimize scheduling strategies using intelligent optimization algorithms; Including multi-objective function construction unit and intelligent optimization algorithm unit; System collaborative scheduling mechanism, used to execute corresponding time-sharing and partitioning scheduling strategies and multi-device collaborative control operations according to different time periods and device status; It includes a time-sharing and partitioning scheduling strategy unit and a multi-device collaborative control unit.

2. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 1 is characterized in that: The photovoltaic module data acquisition unit uses high-precision power sensors, light intensity sensors and temperature sensors to collect the output power, light intensity and temperature data of the photovoltaic module in real time at a frequency of 15 minutes per time; the energy storage device data acquisition unit obtains the charge state, charge and discharge current and voltage data of the energy storage device through the BMS, with a sampling frequency of 5 minutes per time; the motor data acquisition unit uses vibration sensors, current sensors and speed sensors to collect the speed, load and current data of the motor, with a sampling frequency of 1 minute per time; the grid-side data acquisition unit obtains the voltage, frequency and real-time electricity price data of the grid side from the smart meter, with a sampling frequency of 15 minutes per time.

3. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 1 is characterized in that: The data cleaning unit adopts the Z-score method to eliminate abnormal data. The Z-score is obtained by calculating the difference between the data value and the mean and dividing it by the standard deviation. The threshold is pre-set to 3 or -3. When the Z-score of the data exceeds the threshold range, the data is determined to be abnormal data and is eliminated. The data fusion unit adopts the Kalman filter algorithm to realize data fusion, estimates the system state through prediction steps and update steps, and reduces data errors.

4. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 1 is characterized in that: The photovoltaic power generation prediction unit uses a three-layer LSTM network combined with a fully connected layer to output prediction values. The input features include historical photovoltaic power, light intensity, temperature, humidity, and wind speed. The output is a power sequence for the next 1 to 24 hours with a time resolution of 15 minutes. The load demand prediction unit also uses an LSTM neural network. The input features include historical motor load, production plan, and time factor. The output is a load sequence for the next 1 to 12 hours with a time resolution of 15 minutes. The energy storage device aging characteristic prediction unit uses the Arrhenius equation coupling model to predict the effects of temperature and charge and discharge depth on battery aging, and establishes a remaining service life prediction model.

5. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 1 is characterized in that: The multi-objective function construction unit constructs an economic objective function, a reliability objective function, and an environmental benefit objective function, and forms a comprehensive objective function through balancing weight coefficients; the intelligent optimization algorithm unit adopts a particle swarm optimization algorithm, and the particle encoding includes three decision variables: photovoltaic output distribution coefficient, energy storage charging and discharging power, and motor speed regulation ratio. The algorithm process includes initializing the particle swarm, calculating fitness, updating individual optimality and global optimality, speed and position updating, and constraint processing.

6. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 1 is characterized in that: The time-sharing and partitioning scheduling strategy unit executes different scheduling strategies for photovoltaic peak periods, photovoltaic low periods, and transition periods to optimize system operation; the multi-device collaborative control unit is used to handle photovoltaic power mutations and energy storage equipment aging, ensuring that the system can still operate reliably in the face of emergencies and extending the service life of the energy storage equipment.

7. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 6 is characterized in that: During the photovoltaic peak period, photovoltaic power is preferentially allocated to the motor, and the remaining photovoltaic power is used to charge the energy storage device. When the SOC of the energy storage device reaches a fully charged state, the motor speed regulation ratio is intelligently adjusted to increase the motor load and maximize the photovoltaic power consumption. During the photovoltaic off-peak period, based on the load demand forecast results and real-time electricity price information, intelligent decisions are made on discharging the energy storage device and purchasing electricity from the grid. During the low electricity price period, low-priced electricity resources are used to charge the energy storage device. During the transition period, the SOC of the energy storage device and the speed regulation ratio of the motor are adjusted in advance according to the forecast data to smooth out power fluctuations and ensure a smooth transition of the system.

8. The photovoltaic energy storage motor coordinated scheduling optimization system based on multimodal data according to claim 6 is characterized in that: The photovoltaic power mutation processing, when the photovoltaic power drops by more than 20%, triggers the energy storage device's rapid discharge compensation mechanism, and dynamically adjusts the motor speed regulation ratio, reducing the motor speed by 10% to reduce power demand; the energy storage device aging processing, when it is monitored that the remaining service life of the energy storage device is less than 1 year, automatically adjusts the scheduling strategy to limit the DOD to within 50%, reduces the battery aging rate, and extends the service life.

Images