A method and system for intelligent operation of photovoltaic storage integration based on deep learning and multi-objective model predictive control
The photovoltaic-storage integrated system, which utilizes deep learning and multi-objective model predictive control, solves the problem of balancing photovoltaic power generation volatility and battery health, achieving efficient and economical intelligent operation and improving photovoltaic self-consumption rate and battery life.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INSPUR ARTIFICIAL INTELLIGENCE RES INST CO LTD SHANDONG CHINA
- Filing Date
- 2025-12-24
- Publication Date
- 2026-06-19
AI Technical Summary
Existing integrated photovoltaic and energy storage systems cannot simultaneously solve the problem of balancing prediction accuracy and multiple conflicting objectives in environments with high uncertainty, leading to suboptimal economic decisions and neglect of battery health.
A method based on deep learning and multi-objective model predictive control is adopted. Real-time rolling prediction is performed using LSTM and GRU models, combined with a multi-objective optimization function to optimize the charging and discharging strategy of the energy storage system, taking into account economic costs and battery degradation costs, so as to achieve intelligent operation.
It has improved the self-consumption rate of photovoltaic power generation, reduced the total life cycle operating cost, extended battery life, improved system energy efficiency and autonomy, and achieved intelligent integrated operation.
Smart Images
Figure CN121395328B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent control technology for photovoltaic and energy storage systems, specifically relating to an integrated intelligent operation method and system for photovoltaic and energy storage based on deep learning and multi-objective model predictive control. Background Technology
[0002] Existing integrated photovoltaic-storage systems combine photovoltaics, energy storage, and energy management units to improve energy self-sufficiency and economic efficiency. While energy storage systems can mitigate the volatility of photovoltaic power generation and its dependence on weather, they are costly and require efficient operational strategies to ensure economic viability.
[0003] Current mainstream operating methods have the following limitations: Rule-based control methods rely on preset static rules (such as peak-valley arbitrage) for management, which cannot adapt to real-time changes, easily leading to suboptimal economic decisions and neglecting battery health; Offline optimization-based scheduling methods formulate plans based on day-ahead forecasts, but heavily rely on forecast accuracy and cannot cope with real-time deviations, resulting in energy loss or operational imbalances; Conventional Model Predictive Control (MPC) methods use rolling time-domain optimization, but the forecast models are simple, lack accuracy, and have a single optimization objective, focusing only on immediate operating costs without quantifying battery degradation costs. To cope with uncertainty, conservative margins are often set, sacrificing economic benefits. In summary, existing technologies cannot simultaneously solve the problem of forecast accuracy and the challenge of balancing multiple conflicting objectives under high uncertainty environments.
[0004] In view of this, it is very necessary to provide an integrated intelligent operation method and system for photovoltaic and energy storage based on deep learning and multi-objective model predictive control to solve the above-mentioned defects in the prior art. Summary of the Invention
[0005] The purpose of this invention is to address the shortcomings of the existing technology by providing a method and system for the integrated operation of photovoltaic and energy storage based on deep learning and multi-objective model predictive control, thereby solving the aforementioned technical problems.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A method for intelligent operation of integrated photovoltaic and energy storage based on deep learning and multi-objective model predictive control includes the following steps:
[0008] Step S1: At each control time k, use a deep learning model to make real-time rolling predictions of photovoltaic output and user load to obtain the predicted values of photovoltaic output and user load for the next N time steps.
[0009] Step S2: Based on the photovoltaic power output forecast and user load forecast obtained in Step S1, a multi-objective model predictive controller is used to solve a multi-objective optimization problem to obtain the optimal charging and discharging power sequence of the energy storage system. The objective function of the multi-objective optimization problem includes the calculation of economic cost and the calculation of battery degradation cost.
[0010] Step S3: Execute the charge / discharge power command for the first time step in the optimal charge / discharge power sequence obtained in step S2, and repeat steps S1 to S3 at the next control time.
[0011] Preferably, step S1 specifically includes:
[0012] Input historical meteorological data (such as light intensity, temperature, humidity, cloud cover), historical photovoltaic power output data, and timestamp information (such as hour, month) into a photovoltaic power output prediction model based on LSTM (Long Short-Term Memory Network), and output the photovoltaic power output prediction value for the next N time steps.
[0013] Input historical load data, electricity price signals, date type (weekday, weekend, or holiday), and weather data (temperature) into a user load forecasting model based on GRU (Gated Cyclic Unit), and output user load forecast values for the next N time steps;
[0014] The photovoltaic power output forecasting model and the user load forecasting model are executed in a "rolling forecasting" manner: at each control time k (e.g., =09:00) Using the deadline The latest real-time data, generated from arrive (For example, the forecast curve from 09:00 to 09:00 the next day), where The time step is set, and the prediction curve and prediction data are transmitted to the multi-objective model prediction controller.
[0015] Preferably, step S2 specifically includes:
[0016] At each control time k, the multi-objective model predictive controller receives the latest prediction data from the photovoltaic output prediction model and the user load prediction model, and solves for the multi-objective optimization function J. The expression for the multi-objective optimization function J is as follows:
[0017]
[0018] Where k is the current control time, and N is the prediction time domain. The energy storage charging and discharging power sequence is the decision variable for the multi-objective model predictive controller. , and This is a weighting factor, set by the user, used to balance short-term economic benefits with long-term battery life. This range ensures that the controller must take into account the cost of battery degradation during the optimization process. Let be the instantaneous economic cost at time t. Let t be the quantized battery degradation cost at time t.
[0019] Preferably, the instantaneous economic cost and quantified battery degradation cost of the multi-objective optimization function at time t include:
[0020] The instantaneous economic cost at time t is calculated by the economic cost model:
[0021]
[0022] Among them, immediate economic costs This represents the net electricity purchase cost at time t, reflecting the economic efficiency of the interaction between the energy storage system and the power grid; and These are the grid purchase price and sales price of electricity at time t; the grid purchase power. and electricity sales capacity Determined by power balance constraints, satisfying The power balance constraint expression is as follows:
[0023]
[0024] in, Forecasted user load values This is the predicted value for photovoltaic power output. and These are the decision variables for the predictive controller in a multi-objective model.
[0025] The quantized battery degradation cost at time t is calculated using a battery degradation cost model based on energy throughput:
[0026]
[0027] Among them, quantifying battery degradation cost It represents the equivalent economic cost calculated based on the physical life loss (i.e., "cycle aging") caused by charging and discharging the battery at time t. For time step; This is the equivalent degradation cost per unit of energy (kWh) throughput (unit: yuan / kWh), pre-calculated based on the battery's replacement cost and total cycle life (LCC):
[0028]
[0029] in, This is the total replacement cost of the battery system (in yuan). It is the total energy throughput (kWh) over the entire battery life cycle; This refers to the battery's rated capacity (kWh). At the average depth of discharge ( Rated cycle life (cycles) under )
[0030] Multi-objective model predictive controllers only work when the arbitrage space is large enough (i.e. ) or have sufficiently high self-use value (i.e. Charging and discharging only occur when the battery is in a specific condition (e.g., when charging and discharging is not worthwhile). This mechanism avoids those "unworthy," low-value, and damaging micro-charge-discharge cycles, thus intrinsically and intelligently protecting battery life. This is something that existing single-objective optimization techniques cannot achieve.
[0031] Preferably, the multi-objective optimization function solved by the multi-objective model predictive controller further includes:
[0032] To ensure the safe and physically feasible operation of the system, the multi-objective optimization problem-solving process must be carried out at every time step. ( The following constraints must be met:
[0033] (a) Battery SOC state update constraints:
[0034]
[0035] in, For charging efficiency, This refers to the discharge efficiency.
[0036] (b) Battery operating boundary constraints:
[0037] SOC safety upper and lower limit constraints:
[0038] Maximum charging power constraint:
[0039] Maximum discharge power constraint:
[0040] Constraints prohibiting simultaneous charging and discharging:
[0041] (c) Power grid interaction ramp rate constraint:
[0042] To ensure grid-friendliness, the rate of change of grid-injected power and grid-absorbed power is limited to a specified threshold.
[0043] Slope rate constraint:
[0044] Net interactive power constraint:
[0045] in, Net power exchange between the power grid The maximum allowable ramp rate for a power grid is defined as the threshold rate of change of power injected into or absorbed by the grid (e.g., according to relevant technical standards for the Puerto Rican power grid). It can be set to 10% of the system's rated power per minute.
[0046] Furthermore, this invention provides an integrated photovoltaic and energy storage intelligent operation system based on deep learning and multi-objective model predictive control, comprising:
[0047] The AI prediction engine unit, in which:
[0048] At each control time k, a deep learning model is used to make real-time rolling predictions of photovoltaic power output and user load, so as to obtain the predicted values of photovoltaic power output and user load for the next N time steps.
[0049] A multi-objective model predictive controller unit, in which:
[0050] Based on the photovoltaic power output forecast and user load forecast obtained from the AI prediction engine unit, a multi-objective model predictive controller is used to solve a multi-objective optimization problem, resulting in the optimal charging and discharging power sequence of the energy storage system. The objective function of the multi-objective optimization problem includes the calculation of economic cost and the calculation of battery degradation cost.
[0051] The charge / discharge execution unit, in which:
[0052] The system executes the charge / discharge power command for the first time step in the optimal charge / discharge power sequence obtained by the multi-objective model predictive controller unit, and repeats the contents of the AI prediction engine unit, the multi-objective model predictive controller unit, and the charge / discharge execution unit at the next control time step.
[0053] Preferably, the AI prediction engine unit specifically includes:
[0054] Input historical meteorological data (such as light intensity, temperature, humidity, cloud cover), historical photovoltaic power output data, and timestamp information (such as hour, month) into a photovoltaic power output prediction model based on LSTM (Long Short-Term Memory Network), and output the photovoltaic power output prediction value for the next N time steps.
[0055] Input historical load data, electricity price signals, date type (weekday, weekend, or holiday), and weather data (temperature) into a user load forecasting model based on GRU (Gated Cyclic Unit), and output user load forecast values for the next N time steps;
[0056] The photovoltaic power output forecasting model and the user load forecasting model are executed in a "rolling forecasting" manner: at each control time k (e.g., =09:00) Using the deadline The latest real-time data, generated from arrive (For example, the forecast curve from 09:00 to 09:00 the next day), where The time step is set, and the prediction curve and prediction data are transmitted to the multi-objective model prediction controller.
[0057] Preferably, the multi-objective model prediction controller unit specifically includes:
[0058] At each control time k, the multi-objective model predictive controller receives the latest prediction data from the photovoltaic output prediction model and the user load prediction model, and solves for the multi-objective optimization function J. The expression for the multi-objective optimization function J is as follows:
[0059]
[0060] Where k is the current control time, and N is the prediction time domain. The energy storage charging and discharging power sequence is the decision variable for the multi-objective model predictive controller. , and This is a weighting factor, set by the user, used to balance short-term economic benefits with long-term battery life. This range ensures that the controller must take into account the cost of battery degradation during the optimization process. Let be the instantaneous economic cost at time t. Let t be the quantized battery degradation cost at time t.
[0061] Preferably, the multi-objective optimization function includes the instantaneous economic cost and quantified battery degradation cost at time t, comprising:
[0062] The instantaneous economic cost at time t is calculated by the economic cost model:
[0063]
[0064] Among them, immediate economic costs This represents the net electricity purchase cost at time t, reflecting the economic efficiency of the interaction between the energy storage system and the power grid; and These are the grid purchase price and sales price of electricity at time t; the grid purchase power. and electricity sales capacity Determined by power balance constraints, satisfying The power balance constraint expression is as follows:
[0065]
[0066] in, Forecasted user load values This is the predicted value for photovoltaic power output. and These are the decision variables for the predictive controller in a multi-objective model.
[0067] The quantized battery degradation cost at time t is calculated using a battery degradation cost model based on energy throughput:
[0068]
[0069] Among them, quantifying battery degradation cost It represents the equivalent economic cost calculated based on the physical life loss (i.e., "cycle aging") caused by charging and discharging the battery at time t. For time step; This is the equivalent degradation cost per unit of energy (kWh) throughput (unit: yuan / kWh), pre-calculated based on the battery's replacement cost and total cycle life (LCC):
[0070]
[0071] in, This is the total replacement cost of the battery system (in yuan). It is the total energy throughput (kWh) over the entire battery life cycle; This refers to the battery's rated capacity (kWh). At the average depth of discharge ( Rated cycle life (cycles) under )
[0072] Multi-objective model predictive controllers only work when the arbitrage space is large enough (i.e. ) or have sufficiently high self-use value (i.e. Charging and discharging only occur when the battery is in a specific condition (e.g., when charging and discharging is not worthwhile). This mechanism avoids those "unworthy," low-value, and damaging micro-charge-discharge cycles, thus intrinsically and intelligently protecting battery life. This is something that existing single-objective optimization techniques cannot achieve.
[0073] Preferably, the multi-objective optimization function solved in the multi-objective model prediction controller unit further includes:
[0074] To ensure the safe and physically feasible operation of the system, the multi-objective optimization problem-solving process must be carried out at every time step. ( The following constraints must be met:
[0075] (a) Battery SOC state update constraints:
[0076]
[0077] in, For charging efficiency, This refers to the discharge efficiency.
[0078] (b) Battery operating boundary constraints:
[0079] SOC safety upper and lower limit constraints:
[0080] Maximum charging power constraint:
[0081] Maximum discharge power constraint:
[0082] Constraints prohibiting simultaneous charging and discharging:
[0083] (c) Power grid interaction ramp rate constraint:
[0084] To ensure grid-friendliness, the rate of change of grid-injected power and grid-absorbed power is limited to a specified threshold.
[0085] Slope rate constraint:
[0086] Net interactive power constraint:
[0087] in, Net power exchange between the power grid The maximum allowable ramp rate for a power grid is defined as the threshold rate of change of power injected into or absorbed by the grid (e.g., according to relevant technical standards for the Puerto Rican power grid). It can be set to 10% of the system's rated power per minute.
[0088] The beneficial effects of this invention are as follows: First, it reduces the total lifecycle operating cost (LCC) of the energy storage system: Through accurate predictions using LSTM and GRU models, it maximizes energy absorption during actual electricity price troughs (or peak solar power generation) and releases energy during actual electricity price peaks (or peak load periods), resulting in an economic cost optimization effect far exceeding that of static strategies. The battery degradation cost model calculation in the multi-objective optimization function enables the multi-objective prediction model controller to actively extend battery life, avoiding premature battery failure and high replacement costs caused by over-optimizing economic costs. Second, it improves the energy efficiency of the photovoltaic-storage system and features low energy consumption: Based on accurate predictions using LSTM and GRU models, the energy storage system maximizes the "on-site" storage of photovoltaic power and its use when needed, rather than selling it to the grid at a low price or letting it go to waste. This achieves a photovoltaic self-consumption rate of 78%, far exceeding that of rule-based control methods (45% photovoltaic self-consumption rate), thus realizing a "significant improvement in energy performance." By constraining the ramp rate of the objective function of the M-MPC (Multi-Objective Predictive Model Controller), the power curve interacting with the grid is actively smoothed while performing economic dispatch, reducing the impact on the grid. Third, it improves the autonomy and intelligence of the photovoltaic-storage system: through the fully automatic closed-loop control system of this invention, the deep learning model automatically learns and adapts to environmental changes (such as seasonal changes and electricity price adjustments), and the multi-objective predictive model controller automatically seeks optimization. Unlike rule-based control methods that require manual adjustment of peak and off-peak parameters according to seasonal changes, this invention requires no manual intervention, truly achieving "integrated" intelligent operation. Attached Figure Description
[0089] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0090] Figure 1 This is a flowchart of an intelligent operation method for photovoltaic-storage integration based on deep learning and multi-objective model predictive control, provided by the present invention.
[0091] Figure 2 This is a schematic diagram of a photovoltaic-storage integrated intelligent operation system based on deep learning and multi-objective model predictive control, provided by the present invention. Detailed Implementation
[0092] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.
[0093] Example 1:
[0094] like Figure 1 As shown in the figure, this embodiment provides a photovoltaic-storage integrated intelligent operation method based on deep learning and multi-objective model predictive control, which includes the following steps:
[0095] Step S1: At each control time k, use a deep learning model to make real-time rolling predictions of photovoltaic output and user load to obtain the predicted values of photovoltaic output and user load for the next N time steps.
[0096] Step S2: Based on the photovoltaic power output forecast and user load forecast obtained in Step S1, a multi-objective model predictive controller is used to solve a multi-objective optimization problem to obtain the optimal charging and discharging power sequence of the energy storage system. The objective function of the multi-objective optimization problem includes the calculation of economic cost and the calculation of battery degradation cost.
[0097] Step S3: Execute the charge / discharge power command for the first time step in the optimal charge / discharge power sequence obtained in step S2, and repeat steps S1 to S3 at the next control time.
[0098] Step S1 specifically includes:
[0099] Input historical meteorological data (such as light intensity, temperature, humidity, cloud cover), historical photovoltaic power output data, and timestamp information (such as hour, month) into a photovoltaic power output prediction model based on LSTM (Long Short-Term Memory Network), and output the photovoltaic power output prediction value for the next N time steps.
[0100] Input historical load data, electricity price signals, date type (weekday, weekend, or holiday), and weather data (temperature) into a user load forecasting model based on GRU (Gated Cyclic Unit), and output user load forecast values for the next N time steps;
[0101] The photovoltaic power output forecasting model and the user load forecasting model are executed in a "rolling forecasting" manner: at each control time k (e.g., =09:00) Using the deadline The latest real-time data, generated from arrive (For example, the forecast curve from 09:00 to 09:00 the next day), where The time step is set, and the prediction curve and prediction data are transmitted to the multi-objective model prediction controller.
[0102] Step S2 specifically includes:
[0103] At each control time k, the multi-objective model predictive controller receives the latest prediction data from the photovoltaic output prediction model and the user load prediction model, and solves for the multi-objective optimization function J. The expression for the multi-objective optimization function J is as follows:
[0104]
[0105] Where k is the current control time, and N is the prediction time domain. The energy storage charging and discharging power sequence is the decision variable for the multi-objective model predictive controller. , and This is a weighting factor, set by the user, used to balance short-term economic benefits with long-term battery life. This range ensures that the controller must take into account the cost of battery degradation during the optimization process. Let be the instantaneous economic cost at time t. Let t be the quantized battery degradation cost at time t.
[0106] The instantaneous economic cost and quantified battery degradation cost of the multi-objective optimization function at time t in step S2 include:
[0107] The instantaneous economic cost at time t is calculated by the economic cost model:
[0108]
[0109] Among them, immediate economic costs This represents the net electricity purchase cost at time t, reflecting the economic efficiency of the interaction between the energy storage system and the power grid; and These are the grid purchase price and sales price of electricity at time t; the grid purchase power. and electricity sales capacity Determined by power balance constraints, satisfying The power balance constraint expression is as follows:
[0110]
[0111] in, Forecasted user load values This is the predicted value for photovoltaic power output. and These are the decision variables for the predictive controller in a multi-objective model.
[0112] The quantized battery degradation cost at time t is calculated using a battery degradation cost model based on energy throughput:
[0113]
[0114] Among them, quantifying battery degradation cost It represents the equivalent economic cost calculated based on the physical life loss (i.e., "cycle aging") caused by charging and discharging the battery at time t. For time step; This is the equivalent degradation cost per unit of energy (kWh) throughput (unit: yuan / kWh), pre-calculated based on the battery's replacement cost and total cycle life (LCC):
[0115]
[0116] in, This is the total replacement cost of the battery system (in yuan). It is the total energy throughput (kWh) over the entire battery life cycle; This refers to the battery's rated capacity (kWh). At the average depth of discharge ( Rated cycle life (cycles) under )
[0117] Multi-objective model predictive controllers only work when the arbitrage space is large enough (i.e. ) or have sufficiently high self-use value (i.e. Charging and discharging only occur when the battery is in a specific condition (e.g., when charging and discharging is not worthwhile). This mechanism avoids those "unworthy," low-value, and damaging micro-charge-discharge cycles, thus intrinsically and intelligently protecting battery life. This is something that existing single-objective optimization techniques cannot achieve.
[0118] The multi-objective optimization function solved by the multi-objective model predictive controller further includes:
[0119] To ensure the safe and physically feasible operation of the system, the multi-objective optimization problem-solving process must be carried out at every time step. ( The following constraints must be met:
[0120] (a) Battery SOC state update constraints:
[0121]
[0122] in, For charging efficiency, This refers to the discharge efficiency.
[0123] (b) Battery operating boundary constraints:
[0124] SOC safety upper and lower limit constraints:
[0125] Maximum charging power constraint:
[0126] Maximum discharge power constraint:
[0127] Constraints prohibiting simultaneous charging and discharging:
[0128] (c) Power grid interaction ramp rate constraint:
[0129] To ensure grid-friendliness, the rate of change of grid-injected power and grid-absorbed power is limited to a specified threshold.
[0130] Slope rate constraint:
[0131] Net interactive power constraint:
[0132] in, Net power exchange between the power grid The maximum allowable ramp rate for a power grid is defined as the threshold rate of change of power injected into or absorbed by the grid (e.g., according to relevant technical standards for the Puerto Rican power grid). It can be set to 10% of the system's rated power per minute.
[0133] Example 2:
[0134] like Figure 2 As shown in the figure, this embodiment provides a photovoltaic-storage integrated intelligent operation system based on deep learning and multi-objective model predictive control, comprising:
[0135] AI prediction engine unit 1, in which:
[0136] At each control time k, a deep learning model is used to make real-time rolling predictions of photovoltaic power output and user load, so as to obtain the predicted values of photovoltaic power output and user load for the next N time steps.
[0137] Multi-objective model prediction controller unit 2, in which:
[0138] Based on the photovoltaic power output forecast and user load forecast obtained from AI prediction engine unit 1, a multi-objective model prediction controller is used to solve a multi-objective optimization problem to obtain the optimal charging and discharging power sequence of the energy storage system. The objective function of the multi-objective optimization problem includes the calculation of economic cost and the calculation of battery degradation cost.
[0139] Charge / discharge execution unit 3, in which:
[0140] The system executes the charge / discharge power command for the first time step in the optimal charge / discharge power sequence obtained by the multi-objective model prediction controller unit 2, and repeats the contents of the AI prediction engine unit 1, the multi-objective model prediction controller unit 2, and the charge / discharge execution unit 3 at the next control time.
[0141] The AI prediction engine unit 1 specifically includes:
[0142] Input historical meteorological data (such as light intensity, temperature, humidity, cloud cover), historical photovoltaic power output data, and timestamp information (such as hour, month) into a photovoltaic power output prediction model based on LSTM (Long Short-Term Memory Network), and output the photovoltaic power output prediction value for the next N time steps.
[0143] Input historical load data, electricity price signals, date type (weekday, weekend, or holiday), and weather data (temperature) into a user load forecasting model based on GRU (Gated Cyclic Unit), and output user load forecast values for the next N time steps;
[0144] The photovoltaic power output forecasting model and the user load forecasting model are executed in a "rolling forecasting" manner: at each control time k (e.g., =09:00) Using the deadline The latest real-time data, generated from arrive (For example, the forecast curve from 09:00 to 09:00 the next day), where The time step is set, and the prediction curve and prediction data are transmitted to the multi-objective model prediction controller.
[0145] The multi-objective model prediction controller unit 2 specifically includes:
[0146] At each control time k, the multi-objective model predictive controller receives the latest prediction data from the photovoltaic output prediction model and the user load prediction model, and solves for the multi-objective optimization function J. The expression for the multi-objective optimization function J is as follows:
[0147]
[0148] Where k is the current control time, and N is the prediction time domain. The energy storage charging and discharging power sequence is the decision variable for the multi-objective model predictive controller. , and This is a weighting factor, set by the user, used to balance short-term economic benefits with long-term battery life. This range ensures that the controller must take into account the cost of battery degradation during the optimization process. Let be the instantaneous economic cost at time t. Let t be the quantized battery degradation cost at time t.
[0149] The instantaneous economic cost and quantized battery degradation cost of the multi-objective optimization function at time t in the multi-objective model prediction controller unit 2 include:
[0150] The instantaneous economic cost at time t is calculated by the economic cost model:
[0151]
[0152] Among them, immediate economic costs This represents the net electricity purchase cost at time t, reflecting the economic efficiency of the interaction between the energy storage system and the power grid; and These are the grid purchase price and sales price of electricity at time t; the grid purchase power. and electricity sales capacity Determined by power balance constraints, satisfying The power balance constraint expression is as follows:
[0153]
[0154] in, Forecasted user load values This is the predicted value for photovoltaic power output. and These are the decision variables for the predictive controller in a multi-objective model.
[0155] The quantized battery degradation cost at time t is calculated using a battery degradation cost model based on energy throughput:
[0156]
[0157] Among them, quantifying battery degradation cost It represents the equivalent economic cost calculated based on the physical life loss (i.e., "cycle aging") caused by charging and discharging the battery at time t. For time step; This is the equivalent degradation cost per unit of energy (kWh) throughput (unit: yuan / kWh), pre-calculated based on the battery's replacement cost and total cycle life (LCC):
[0158]
[0159] in, This is the total replacement cost of the battery system (in yuan). It is the total energy throughput (kWh) over the entire battery life cycle; This refers to the battery's rated capacity (kWh). At the average depth of discharge ( Rated cycle life (cycles) under )
[0160] Multi-objective model predictive controllers only work when the arbitrage space is large enough (i.e. ) or have sufficiently high self-use value (i.e. Charging and discharging only occur when the battery is in a specific condition (e.g., when charging and discharging is not worthwhile). This mechanism avoids those "unworthy," low-value, and damaging micro-charge-discharge cycles, thus intrinsically and intelligently protecting battery life. This is something that existing single-objective optimization techniques cannot achieve.
[0161] The multi-objective optimization function solved in the multi-objective model prediction controller unit 2 also includes:
[0162] To ensure the safe and physically feasible operation of the system, the multi-objective optimization problem-solving process must be carried out at every time step. ( The following constraints must be met:
[0163] (a) Battery SOC state update constraints:
[0164]
[0165] in, For charging efficiency, This refers to the discharge efficiency.
[0166] (b) Battery operating boundary constraints:
[0167] SOC safety upper and lower limit constraints:
[0168] Maximum charging power constraint:
[0169] Maximum discharge power constraint:
[0170] Constraints prohibiting simultaneous charging and discharging:
[0171] (c) Power grid interaction ramp rate constraint:
[0172] To ensure grid-friendliness, the rate of change of grid-injected power and grid-absorbed power is limited to a specified threshold.
[0173] Slope rate constraint:
[0174] Net interactive power constraint:
[0175] in, Net power exchange between the power grid The maximum allowable ramp rate for a power grid is defined as the threshold rate of change of power injected into or absorbed by the grid (e.g., according to relevant technical standards for the Puerto Rican power grid). It can be set to 10% of the system's rated power per minute.
[0176] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The methods disclosed in the embodiments are described simply because they correspond to the systems disclosed in the embodiments; relevant details can be found in the method section.
[0177] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0178] In the embodiments provided by this invention, it should be understood that the disclosed systems, methods, and approaches can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.
[0179] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0180] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit.
[0181] Similarly, in the various embodiments of the present invention, each processing unit can be integrated into a functional module, or each processing unit can exist physically, or two or more processing units can be integrated into a functional module.
[0182] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0183] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0184] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.
Claims
1. A method for intelligent operation of photovoltaic storage integration based on deep learning and multi-objective model predictive control, characterized in that, Includes the following steps: Step S1: At each control time k, use a deep learning model to make real-time rolling predictions of photovoltaic output and user load to obtain the predicted values of photovoltaic output and user load for the next N time steps. Step S2: Based on the photovoltaic power output forecast and user load forecast obtained in Step S1, a multi-objective model predictive controller is used to solve a multi-objective optimization problem to obtain the optimal charging and discharging power sequence of the energy storage system. The objective function of the multi-objective optimization problem includes the calculation of economic cost and the calculation of battery degradation cost. Step S3: Execute the charge / discharge power command for the first time step in the optimal charge / discharge power sequence obtained in step S2, and repeat steps S1 to S3 at the next control time. Step S1 specifically includes: Input historical meteorological data, historical photovoltaic power output data, and timestamp information into the LSTM-based photovoltaic power output prediction model, and output the photovoltaic power output prediction value for the next N time steps; Input historical load data, electricity price signals, date type, and weather data into the GRU-based user load forecasting model, and output the user load forecast values for the next N time steps; The photovoltaic output forecasting model and the user load forecasting model are executed in a rolling forecasting manner: at each control time k, the cutoff date is used. The latest real-time data, generated from arrive The predicted curve, where The time step is set, and the prediction curve and prediction data are transmitted to the multi-objective model prediction controller; Step S2 specifically includes: At each control time k, the multi-objective model predictive controller receives the latest prediction data from the photovoltaic power output prediction model and the user load prediction model, and solves the multi-objective optimization function J; the expression of the multi-objective optimization function J is as follows: Where k is the current control time, and N is the prediction time domain. The energy storage charging and discharging power sequence is the decision variable for the multi-objective model predictive controller. , and This is a weighting factor, set by the user, used to balance short-term economic benefits with long-term battery life. ; Let be the instantaneous economic cost at time t. Let be the quantized battery degradation cost at time t; The instantaneous economic cost and quantified battery degradation cost of the multi-objective optimization function at time t include: The instantaneous economic cost at time t is calculated by the economic cost model: Among them, immediate economic costs This represents the net cost of electricity purchased at time t; and These are the grid purchase price and sales price of electricity at time t; the grid purchase power. and electricity sales capacity Determined by power balance constraints; The quantized battery degradation cost at time t is calculated using a battery degradation cost model based on energy throughput: Among them, quantifying battery degradation cost This represents the equivalent economic cost calculated based on the physical lifespan loss caused by charging and discharging the battery at time t. For time step, and For predicting the decision variables of the controller in a multi-objective model, It is the equivalent decay cost per unit of energy throughput; The equivalent attenuation cost per unit of energy throughput Based on the battery's replacement cost and total cycle life, the following calculations were performed in advance: wherein, is the total replacement cost of the battery system; is the total energy throughput of the battery over its full life cycle; is the rated capacity of the battery; is the rated cycle life at an average depth of discharge of 50%. 2.The photovoltaic storage integrated intelligent operation method based on deep learning and multi-objective model predictive control according to claim 1, wherein, The multi-objective optimization function solved by the multi-objective model predictive controller also includes: The solution process of the multi-objective optimization problem must satisfy the following constraints at each time instant the following constraints: (a) Battery SOC state update constraint: Ensure that the battery's state of charge at each moment conforms to the recursive relationship based on charge and discharge power and efficiency, and that the recursive update process is matched with the charging power, discharging power, charging efficiency, discharging efficiency and rated capacity. (b) Battery operating boundary constraints: including SOC safety upper and lower limits constraints, maximum charge and discharge power constraints, and operating logic constraints that prohibit simultaneous charge and discharge; (c) Grid interaction ramp rate constraint: The variation of the net grid interaction power between adjacent time points is limited to the maximum allowable ramp rate of the grid. The net grid interaction power is composed of purchased power and sold power.
3. A photovoltaic-storage integrated intelligent operation system based on deep learning and multi-objective model predictive control, characterized in that, include: The AI prediction engine unit, in which: At each control time k, a deep learning model is used to make real-time rolling predictions of photovoltaic power output and user load, so as to obtain the predicted values of photovoltaic power output and user load for the next N time steps. A multi-objective model predictive controller unit, in which: Based on the photovoltaic power output forecast and user load forecast obtained from the AI prediction engine unit, a multi-objective model predictive controller is used to solve a multi-objective optimization problem, resulting in the optimal charging and discharging power sequence of the energy storage system. The objective function of the multi-objective optimization problem includes the calculation of economic cost and the calculation of battery degradation cost. The charge / discharge execution unit, in which: Execute the charge and discharge power command for the first time step in the optimal charge and discharge power sequence obtained by the multi-objective model prediction controller unit, and repeat the contents of the AI prediction engine unit, the multi-objective model prediction controller unit, and the charge and discharge execution unit at the next control time. The AI prediction engine unit specifically includes: Input historical meteorological data, historical photovoltaic power output data, and timestamp information into the LSTM-based photovoltaic power output prediction model, and output the photovoltaic power output prediction value for the next N time steps; Input historical load data, electricity price signals, date type, and weather data into the GRU-based user load forecasting model, and output the user load forecast values for the next N time steps; The photovoltaic output forecasting model and the user load forecasting model are executed in a rolling forecasting manner: at each control time k, the cutoff date is used. The latest real-time data, generated from arrive The prediction curve, in which The time step is set, and the prediction curve and prediction data are transmitted to the multi-objective model prediction controller; The multi-objective model prediction controller unit specifically includes: At each control time k, the multi-objective model predictive controller receives the latest prediction data from the photovoltaic power output prediction model and the user load prediction model, and solves the multi-objective optimization function J; the expression of the multi-objective optimization function J is as follows: Where k is the current control time, and N is the prediction time domain. The energy storage charging and discharging power sequence is the decision variable for the multi-objective model predictive controller. , and This is a weighting factor, set by the user, used to balance short-term economic benefits with long-term battery life. ; Let be the instantaneous economic cost at time t. Let be the quantized battery degradation cost at time t; The instantaneous economic cost and quantified battery degradation cost of the multi-objective optimization function at time t include: The instantaneous economic cost at time t is calculated by the economic cost model: Among them, immediate economic costs This represents the net cost of electricity purchased at time t; and These are the grid purchase price and sales price of electricity at time t; the grid purchase power. and electricity sales capacity Determined by power balance constraints; The quantized battery degradation cost at time t is calculated using a battery degradation cost model based on energy throughput: Among them, quantifying battery degradation cost This represents the equivalent economic cost calculated based on the physical lifespan loss caused by charging and discharging the battery at time t. For time step; It is the equivalent decay cost per unit of energy throughput; and For predicting the decision variables of the controller in a multi-objective model; The equivalent attenuation cost per unit of energy throughput Based on the battery's replacement cost and total cycle life, the following calculations were performed in advance: in, It is the total replacement cost of the battery system; It is the total energy throughput of the battery throughout its entire life cycle; This refers to the battery's rated capacity. In the average depth of discharge The rated cycle life below.
4. The integrated photovoltaic and energy storage intelligent operation system based on deep learning and multi-objective model predictive control according to claim 3, characterized in that, The multi-objective optimization function solved in the multi-objective model predictive controller unit further includes: The solution process for multi-objective optimization problems must be carried out at every time step. The following constraints must be met: (a) Battery SOC state update constraint: Ensure that the battery's state of charge at each moment conforms to the recursive relationship based on charge and discharge power and efficiency, and that the recursive update process is matched with the charging power, discharging power, charging efficiency, discharging efficiency and rated capacity. (b) Battery operating boundary constraints: including SOC safety upper and lower limits constraints, maximum charge and discharge power constraints, and operating logic constraints that prohibit simultaneous charge and discharge; (c) Grid interaction ramp rate constraint: The change in net grid interaction power between adjacent time points is limited to the maximum allowable ramp rate. Net grid interaction power consists of purchased power and sold power.