A method and system for optimizing operation of an energy storage system taking into account dynamic electricity price signals
Patent Information
- Application Number
- CN202611486666.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-23
- Publication Date
- 2026-10-27
AI Technical Summary
[0002]现有储能调控运行方法主要分为两类：一类是基于固定分时电价的储能充放电策略，通过预设的分时电价制定充放电计划，该类方法难以适应动态电价信号的快速变化特征，无法充分捕捉现货市场的价格波动收益；另一类是基于预测与优化的调度方法，通过对负荷、光伏及电价进行预测，建立以用电成本最小化为目标的数学规划模型，采用混合整数线性规划、二次规划等算法求解最优充放电计划
1、本方法将动态电价信号转化为荷电状态偏移量，通过虚拟阈值比较直接生成充放电决策，无需在线求解混合整数规划或二次规划等复杂优化问题，降低了计算资源需求，可在资源受限的园区边缘计算设备上部署实施；
Smart Images
Figure CN122992943A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of energy dispatching technology, and in particular to a method and system for optimizing the operation of an energy storage system that takes into account dynamic electricity price signals. Background Technology
[0002] Existing energy storage regulation and operation methods are mainly divided into two categories: one is the energy storage charging and discharging strategy based on fixed time-of-use electricity prices, which formulates charging and discharging plans through preset time-of-use electricity prices. This type of method is difficult to adapt to the rapid changes in dynamic electricity price signals and cannot fully capture the gains from price fluctuations in the spot market; the other is the scheduling method based on prediction and optimization, which establishes a mathematical programming model with the goal of minimizing electricity costs by predicting load, photovoltaics, and electricity prices, and uses algorithms such as mixed integer linear programming and quadratic programming to solve for the optimal charging and discharging plan.
[0003] However, existing prediction and optimization-based energy storage control methods still have the following shortcomings: First, the optimization models are highly complex and computationally intensive, requiring high-performance computing platforms and making them difficult to deploy in resource-constrained edge devices within industrial parks. Second, the multi-timescale coordination mechanisms are complex, with the connection between day-ahead planning and intraday rolling adjustments often relying on complex reference trajectory tracking or soft constraint penalties, making parameter tuning difficult. Third, the electricity price signal is only used as an external input to the optimization objective function and fails to form an organic linkage with the operating status of the energy storage system; the response speed of the control strategy to changes in electricity price is limited by the optimization solution cycle. Therefore, there is an urgent need for a computationally simple, fast-responding energy storage system control and operation method that can deeply integrate the energy storage operating status with dynamic electricity price signals. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides a method and system for optimizing the operation of energy storage systems that takes into account dynamic electricity price signals.
[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals. The method includes the following steps: acquiring load power, energy storage system operating status data, and photovoltaic power generation data of an industrial park, and simultaneously acquiring electricity price and meteorological data to construct a historical dataset; based on the historical dataset, acquiring day-ahead load forecast results, photovoltaic power generation forecast results, and electricity price forecast results, as well as short-term intraday forecast results of load and electricity price; determining control parameters for intraday regulation based on the day-ahead forecast results, combined with the rated capacity and state-of-charge (SOC) operating range of the energy storage system; generating a charging and discharging power command for the energy storage system using the control parameters, the electricity price forecast results, the SOC of the energy storage system, and the short-term forecast results; and performing closed-loop correction and safety limiting on the charging and discharging power command based on the actual output power of the energy storage converter and the SOC of the energy storage system before issuing it to the energy storage converter for execution. This invention forms a virtual state of charge control mechanism by embedding dynamic electricity price signals into the energy storage state of charge. It enables the energy storage system to respond quickly to dynamic electricity prices without the need for complex optimization solutions, and combines computational simplicity with control effectiveness.
[0006] Optionally, the energy storage system operating status data includes the energy storage system's state of charge and the actual output power of the energy storage converter; the photovoltaic power generation data is the photovoltaic output power of the photovoltaic power generation system; and the meteorological data includes actual meteorological data and meteorological forecast data, including temperature forecast data, humidity forecast data, wind speed forecast data, and sunshine forecast data.
[0007] Optionally, the steps of acquiring the load power, energy storage system operation status data, and photovoltaic power generation data of the industrial park, while simultaneously acquiring electricity price and meteorological data, and then constructing a historical dataset, include the following: Real-time collection of raw load power, raw energy storage system operation status data, and raw photovoltaic power generation data of industrial parks, and simultaneous acquisition of raw electricity prices and raw meteorological data; Preprocessing is performed on the raw load power, the raw energy storage system operating status data, the raw photovoltaic power generation data, the raw electricity price, and the raw meteorological data. A historical dataset is constructed using the preprocessed load power, the energy storage system operating status data, the photovoltaic power generation data, the electricity price, and the weather forecast data.
[0008] Optionally, the step of obtaining day-ahead load forecasting results, photovoltaic power generation forecasting results, and electricity price forecasting results based on the historical dataset, as well as short-term intraday load and electricity price forecasting results, includes the following steps: A prediction model is constructed based on the historical dataset. The prediction model includes a load prediction model, a photovoltaic power generation prediction model, an electricity price prediction model, and an intraday short-term prediction model. The load power in the recent period, the actual meteorological data, and the meteorological forecast data for the next day are input into the load prediction model to obtain the load prediction results at the day-ahead scale. The photovoltaic output power in the recent period, the actual meteorological data, and the meteorological forecast data for the next day are input into the photovoltaic power generation prediction model to obtain the photovoltaic power generation prediction results on a day-ahead scale. The electricity price over a recent period is input into the electricity price forecasting model to obtain the day-ahead electricity price forecast result; The load power, electricity price, actual meteorological data, and future short-term meteorological forecast data are input into the intraday short-term prediction model to obtain the short-term prediction results.
[0009] Optionally, the step of constructing a prediction model based on the historical dataset, wherein the prediction model includes a load prediction model, a photovoltaic power generation prediction model, an electricity price prediction model, and an intraday short-term prediction model, includes the following steps: Based on the aforementioned historical dataset, the first dataset, the second dataset, the third dataset, and the fourth dataset are constructed respectively. The LSTM was trained and validated using the first dataset, the second dataset, the third dataset, and the fourth dataset, respectively, to obtain the load forecasting model, the photovoltaic power generation forecasting model, the electricity price forecasting model, and the intraday short-term forecasting model.
[0010] Optionally, the control parameters include a benchmark electricity price, an electricity price sensitivity coefficient, an upper limit for virtual state of charge, and an upper limit for virtual state of charge.
[0011] Optionally, determining the control parameters for intraday regulation based on the day-ahead forecast results, combined with the rated capacity and state-of-charge operating range of the energy storage system, includes the following steps: The load forecast results are summed and normalized, and the normalized load forecast results are used as the weights of the electricity price at the corresponding time. The weighted sum of the electricity price forecast results within the day is used as the benchmark electricity price. The maximum daily electricity price difference is calculated based on the electricity price forecast results, and then the ratio of the rated capacity to the maximum daily electricity price difference is calculated. The electricity price sensitivity coefficient is determined in combination with the safety margin coefficient. The benchmark electricity price and the electricity price sensitivity coefficient are updated using the short-term forecast results according to a preset time step. The virtual state of charge upper limit and the virtual state of charge lower limit are determined using the state of charge operating range of the energy storage system.
[0012] Optionally, generating the charging and discharging power command for the energy storage system using the control parameters, the electricity price forecast result, the state of charge of the energy storage system, and the short-term forecast result includes the following steps: Using the aforementioned control parameters, the short-term prediction results, and the energy storage system's state of charge, the virtual state of charge of the energy storage system is calculated in real time. When the virtual state of charge is greater than or equal to the upper limit of the virtual state of charge, the energy storage system discharges, and the discharge power is calculated in combination with the rated power of the energy storage converter to form a discharge power command; When the virtual state of charge is less than or equal to the virtual state of charge limit, the energy storage system charges, and the charging power is calculated in combination with the rated power of the energy storage converter to form a charging power command.
[0013] Optionally, the step of performing closed-loop correction and safety limiting on the charging and discharging power command based on the actual output power of the energy storage converter and the state of charge of the energy storage system before issuing it to the energy storage converter for execution includes the following steps: The power deviation between the charging and discharging power command and the actual output power of the energy storage converter is calculated, and then the charging and discharging power command is corrected in a closed loop through PI control. The modified charging and discharging power command is safely limited based on the state of charge of the energy storage system to obtain the final power command, which is then sent to the energy storage converter for execution.
[0014] Secondly, the present invention provides an energy storage system operation optimization system that takes into account dynamic electricity price signals. The energy storage system operation optimization system that takes into account dynamic electricity price signals includes: a data acquisition device, a data output device, a processor, and a storage device. The storage device includes a computer-readable storage medium storing a computer program. The computer program includes program instructions. When the program instructions are executed by the processor, the processor enables the processor to implement the energy storage system operation optimization method that takes into account dynamic electricity price signals provided by the present invention.
[0015] In summary, the present invention has at least the following beneficial effects: 1. This method converts dynamic electricity price signals into state of charge offsets and directly generates charging and discharging decisions through virtual threshold comparisons. It eliminates the need for online solving of complex optimization problems such as mixed integer programming or quadratic programming, reducing the computational resource requirements and enabling deployment on resource-constrained edge computing devices in campuses. 2. This method links the virtual state of charge with the actual state of charge in real time. Changes in electricity price can be reflected in changes in the virtual state of charge immediately. Charging and discharging decisions do not need to wait for the optimization solution cycle. It has a fast response speed to dynamic electricity price signals and can fully capture the price fluctuation opportunities in the spot market. 3. This method ensures overall economic efficiency within the scheduling cycle by determining the control parameters at the day-ahead scale, dynamically adjusts the intraday scale based on ultra-short-term forecasts to adapt to source-load fluctuations, and ensures control accuracy by real-time power correction, thus achieving an effective balance between long-term planning and short-term adaptation. 4. This method uses the actual state of charge as the basis for judgment to impose hard constraints on charging and discharging commands, ensuring that the energy storage system always operates within the safe operating range, effectively preventing overcharging and over-discharging, and ensuring battery safety and stable system operation. 5. A system adapted to the method is provided, which not only improves the practicality of the method but also facilitates its promotion. Attached Figure Description
[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating an energy storage system operation optimization method that takes into account dynamic electricity price signals, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the framework of an energy storage system operation optimization system that takes into account dynamic electricity price signals, according to an embodiment of the present invention. Detailed Implementation
[0018] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.
[0019] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.
[0020] It should be noted in advance that, in one alternative embodiment, except for independent descriptions, the same symbols or letters appearing in all formulas have the same meaning.
[0021] In one optional embodiment, please refer to Figure 1 This invention provides a method for optimizing the operation of an energy storage system that takes into account dynamic electricity price signals. The method includes the following steps: S1. Acquire data on the load power, energy storage system operation status, and photovoltaic power generation of the industrial park, and simultaneously acquire electricity price and meteorological data to build a historical dataset.
[0022] The energy storage system's operational status data includes the energy storage system's state of charge and the actual output power of the energy storage converter. Photovoltaic power generation data refers to the photovoltaic output power of the photovoltaic power generation system. Meteorological data includes actual meteorological data and weather forecast data. Actual meteorological data includes actual temperature data, actual humidity data, actual wind speed data, and actual sunshine data. Weather forecast data includes temperature forecast data, humidity forecast data, wind speed forecast data, and sunshine forecast data. Step S1 specifically includes the following steps: S11. Real-time collection of raw load power, raw energy storage system operation status data, and raw photovoltaic power generation data of the industrial park, and simultaneous acquisition of raw electricity price and raw meteorological data.
[0023] Specifically, in this embodiment, the raw energy storage system operating status data includes the raw energy storage system state of charge (i.e., the actual state of charge of the energy storage system directly collected) and the raw energy storage converter actual output power (i.e., the actual switching power of the AC side of the energy storage converter directly collected). The raw photovoltaic power generation data is the raw photovoltaic output power of the photovoltaic power generation system. The raw meteorological data includes raw actual meteorological data and raw meteorological forecast data. The raw actual meteorological data includes raw actual temperature data, raw actual humidity data, raw actual wind speed data, and raw actual sunshine data. The raw meteorological forecast data includes raw temperature forecast data, raw humidity forecast data, raw wind speed forecast data, and raw sunshine forecast data for the next day. The collected raw load power, raw energy storage system operating status data, raw photovoltaic power generation data, raw electricity price, and raw meteorological data are stored in the data buffer of the energy management platform of the industrial park for subsequent data preprocessing.
[0024] More specifically, the raw load power is collected in real time by meters installed at the main incoming line (high-voltage switchgear or low-voltage switchgear) of the industrial park. On the energy storage system side, the battery management system obtains the raw state of charge of the energy storage system in real time, and the controller of the energy storage converter collects the actual output power of the energy storage converter. On the photovoltaic power generation system side of the industrial park, the raw photovoltaic output power of the photovoltaic power generation system is collected in real time through the communication interface of the photovoltaic inverter or combiner box, i.e., the raw real-time output power of the photovoltaic power generation system. All of the above data is transmitted to the energy management platform of the industrial park via industrial Ethernet or RS485 communication link.
[0025] The raw electricity price is obtained through the data interface between the industrial park's energy management platform and the power trading center, using the HTTP / HTTPS communication protocol. Specifically, it is the real-time market price, and the acquisition frequency is consistent with the market clearing frequency. Raw actual temperature, humidity, wind speed, and illumination data are directly collected by temperature, humidity, wind speed, and illumination sensors deployed throughout the industrial park. The data collected by each sensor is transmitted to the energy management platform via an RS485 communication link. Raw weather forecast data is obtained directly through the data interface between the industrial park's energy management platform and a third-party meteorological service provider.
[0026] S12. Preprocess the original load power, the original energy storage system operation status data, the original photovoltaic power generation data, the original electricity price, and the original meteorological data.
[0027] Specifically, in this embodiment, the preprocessing includes outlier identification and removal, missing value filling, physical threshold limiting, resampling, and normalization. For each type of data—original load power, original energy storage system state of charge, original energy storage converter actual output power, original photovoltaic output power, original electricity price, original actual temperature data, original actual humidity data, original actual wind speed data, original actual sunshine data, original temperature forecast data, original humidity forecast data, original wind speed forecast data, and original sunshine forecast data—outlier identification and removal are first performed based on the Raida criterion, and then missing values are filled using forward filling.
[0028] After completing the missing value filling, physical threshold limits are applied to the original load power, original energy storage system state of charge, original energy storage converter actual output power, and original photovoltaic output power. These limits stipulate that their values should not be lower than the corresponding lower limit and should not exceed the corresponding upper limit. For any one of these data points—original load power, original energy storage system state of charge, original energy storage converter actual output power, and original photovoltaic output power—if it is lower than the corresponding lower limit at any given time, its value at that time is modified to the corresponding lower limit; if it is higher than the corresponding upper limit at any given time, its value at that time is modified to the corresponding upper limit. The lower limit for the original load power is 0, and the upper limit is [missing value]. , The rated capacity of the main transformer in the industrial park is specified. This main transformer is typically the one connecting the power grid (or the upstream substation) to the internal distribution network of the industrial park. The lower limit of the initial energy storage system's state of charge is 0%, and the upper limit is 100%. The lower limit of the actual output power of the initial energy storage converter is specified. The upper limit is , The rated power of the energy storage converter is indicated by "+" for discharging and "-" for charging; the lower limit of the original photovoltaic output power is 0, and the upper limit is... , This refers to the installed capacity of the photovoltaic power generation system.
[0029] Furthermore, a linear interpolation method was used to resample each type of data to the same sampling frequency. Then, a maximum-minimum normalization method was applied to normalize each type of data, yielding the final load power, energy storage system operating status data, photovoltaic power generation data, electricity price, and weather forecast data. The time interval between two adjacent sampling times after resampling was 15 minutes.
[0030] S13. Construct a historical dataset using the preprocessed load power, the energy storage system operating status data, the photovoltaic power generation data, the electricity price, and the weather forecast data.
[0031] Specifically, in this embodiment, each type of data, including historical load power, energy storage system operating status data, photovoltaic power generation data, electricity price and weather forecast data, is stored in a time series manner to obtain a historical dataset.
[0032] S2. Based on the historical dataset, obtain the day-ahead load forecast, photovoltaic power generation forecast, and electricity price forecast, as well as the day-ahead short-term forecast of load and electricity price.
[0033] The term "before the day" refers to the period before the forecast period begins, but before the forecast period has commenced. The term "within the day" refers to the period after the forecast period has commenced. Step S2 specifically includes the following steps: S21. Construct a prediction model based on the historical dataset. The prediction model includes a load prediction model, a photovoltaic power generation prediction model, an electricity price prediction model, and an intraday short-term prediction model.
[0034] Step S21 specifically includes the following steps: S211. Construct the first dataset, the second dataset, the third dataset, and the fourth dataset based on the historical dataset.
[0035] Specifically, in this embodiment, the load power and actual meteorological data obtained over four consecutive days are used as a first type of sample data, and multiple first type sample data are extracted from the historical dataset to construct a first dataset; the photovoltaic output power and actual meteorological data obtained over four consecutive days are used as a second type of sample data, and multiple second type sample data are extracted from the historical dataset to construct a second dataset; the electricity price obtained over four consecutive days is used as a third type of sample data, and multiple third type sample data are extracted from the historical dataset to construct a third dataset; the load power, electricity price and actual meteorological data obtained over 12 consecutive hours are used as a fourth type of sample data, and multiple fourth type sample data are extracted from the historical dataset to construct a fourth dataset.
[0036] S212. Use the first dataset, the second dataset, the third dataset, and the fourth dataset to train and validate the LSTM, respectively, to obtain the load forecasting model, the photovoltaic power generation forecasting model, the electricity price forecasting model, and the intraday short-term forecasting model.
[0037] Specifically, in this embodiment, a load forecasting network, a photovoltaic power generation forecasting network, an electricity price forecasting network, and an intraday short-term forecasting network are first constructed based on a Sequence-to-Sequence Long Short-Term Memory (Seq2Seq LSTM) network. The input to the load forecasting network can be represented as... , This provides the load power and actual meteorological data for the i-th sampling time within the three days prior to the next day. For the actual meteorological data at the j-th sampling time within the next day, , , The load power at the i-th sampling time within the next three days. This refers to the actual temperature data at the i-th sampling time within the three days prior to the next day. This refers to the actual humidity data at the i-th sampling time within the three days prior to the next day. The data represents the actual wind speed at the i-th sampling time within the three days prior to the next day. Let N be the actual illumination data at the i-th sampling time within the three days prior to the next day, and N be the total number of sampling times within the three days prior to the next day. For the sampling data at the j-th sampling time within the next day, , This refers to the actual temperature data at the j-th sampling time within the next day. This refers to the actual humidity data at the j-th sampling time within the next day. For the actual wind speed data at the j-th sampling time within the next day, This refers to the actual illumination data at the j-th sampling time within the next day. M represents the total number of sampling times within the next day.
[0038] In terms of network structure, the load prediction network's encoder uses a bidirectional long short-term memory (BiLSTM) network with 64 neurons, and sets `return_sequences=True` to retain the hidden state at each time step. A dropout layer with a dropout rate of 0.2 follows the encoder output to suppress overfitting. The load prediction network's decoder uses a unidirectional long short-term memory (LSTM) network, also with 64 neurons. It receives the context vector from the encoder output as its initial state and generates load power for each sampling time step within the next day. A dropout layer with a dropout rate of 0.2 follows the decoder, and finally, a fully connected layer and a linear activation function map the decoder output to a one-dimensional load power prediction value.
[0039] The input to the photovoltaic power generation prediction network can be expressed as: , This provides the photovoltaic output power and actual meteorological data for the i-th sampling time within the three days prior to the next day. , This represents the photovoltaic (PV) output power at the i-th sampling time within the three days prior to the next day. The encoder of the PV power generation prediction network uses a BiLSTM with 128 neurons, and sets `return_sequences=True` to preserve the hidden state at each time step. It is followed by a Dropout layer with a dropout rate of 0.3. The decoder of the PV power generation prediction network uses a unidirectional LSTM, also with 128 neurons, to decode and generate the PV output power for each sampling time within the next day. The decoder is also followed by a Dropout layer with a dropout rate of 0.3. Finally, a fully connected layer and a linear activation function map the decoder output to a one-dimensional PV output power prediction value.
[0040] For an electricity price forecasting network, its input can be represented as: , The electricity price prediction network is for the electricity price at the i-th sampling time within the three days prior to the next day. The encoder uses a unidirectional LSTM with 50 neurons, and `return_sequences=True`. A dropout layer with a dropout rate of 0.2 follows the encoder. The decoder also uses a unidirectional LSTM with 50 neurons to decode and generate the electricity price for each sampling time within the next day. A dropout layer with a dropout rate of 0.2 follows the decoder, and finally, a fully connected layer and a linear activation function map the decoder output to a one-dimensional electricity price prediction value.
[0041] The input to the intraday short-term forecasting network can be represented as: , This provides the load power, electricity price, and actual meteorological data for the r-th sampling time within the preceding 8 hours (4 hours to come). , , This represents the load power at the r-th sampling time within the previous 8 hours of the next 4 hours. This refers to the actual temperature data at the r-th sampling time within the previous 8 hours of the next 4 hours. This refers to the actual humidity data at the r-th sampling time within the previous 8 hours (4 hours prior to the next). This refers to the actual wind speed data at the r-th sampling time within the previous 8 hours (4 hours in total). This represents the actual illumination data at the r-th sampling time within the previous 8 hours of the next 4 hours, where n is the total number of sampling times within the previous 8 hours of the next 4 hours. This represents the actual meteorological data at the k-th sampling time within the next 4 hours. , This is the actual temperature data at the k-th sampling time within the next 4 hours. This represents the actual humidity data at the k-th sampling time within the next 4 hours. This represents the actual wind speed data at the k-th sampling time within the next 4 hours. This represents the actual illumination data at the k-th sampling time within the next 4 hours. , where m is the total number of sampling times in the next 4 hours.
[0042] In terms of network structure, considering that ultra-short-term prediction relies more on the local features of recent sequences, the encoder of the intraday short-term prediction network uses a unidirectional LSTM with 128 neurons to ensure sufficient nonlinear fitting capability, and returns_sequences=True. Simultaneously, a dropout layer with a dropout rate of 0.2 follows the encoder. The decoder of the intraday short-term prediction network uses a unidirectional LSTM with 128 neurons, receiving the context vector output from the encoder to decode and generate load power and electricity price for each sampling time in the next 4 hours. A dropout layer with a dropout rate of 0.2 follows the decoder, and finally, a dual-output fully connected layer and a linear activation function map the decoder output to intraday short-term load power and electricity price predictions.
[0043] The first dataset was divided into training and validation sets in a 7:3 ratio to train and validate the load prediction network, resulting in the load prediction model. During training, the mean squared error between the predicted and actual load power was used as the loss function. The Adam optimizer was used, with an initial learning rate of 0.001. To prevent premature overfitting, early stopping was employed during training; training was terminated when the loss of the load prediction network on the validation set stopped decreasing after 20 consecutive training epochs.
[0044] Referring to the method of obtaining the load forecasting model based on the first dataset, the second dataset was used to complete the training and validation of the photovoltaic power generation forecasting network, and the photovoltaic power generation forecasting model was obtained; the third dataset was used to complete the training and validation of the electricity price forecasting network, and the electricity price forecasting model was obtained; the fourth dataset was used to complete the training and validation of the intraday short-term forecasting network, and the intraday short-term forecasting model was obtained.
[0045] S22. Input the load power in the recent period, the actual meteorological data, and the meteorological forecast data for the next day into the load prediction model to obtain the load prediction result at the day-ahead scale.
[0046] Specifically, in this embodiment, the load power of the most recent 3 days, the actual meteorological data, and the meteorological forecast data for the next day are input into the load prediction model to obtain the load power prediction value for each sampling time within the next day. Then, the load power prediction value for each sampling time is denormalized to obtain the load prediction result for the corresponding sampling time within the next day.
[0047] S23. Input the photovoltaic output power in the recent period, the actual meteorological data, and the meteorological forecast data for the next day into the photovoltaic power generation prediction model to obtain the photovoltaic power generation prediction result on a day-ahead scale.
[0048] Specifically, in this embodiment, the photovoltaic output power of the most recent 3 days, the actual meteorological data, and the meteorological forecast data for the next day are input into the photovoltaic power generation prediction model to obtain the photovoltaic output power prediction value for each sampling time within the next day. Then, the photovoltaic output power prediction value for each sampling time is denormalized to obtain the photovoltaic power generation prediction result for the corresponding sampling time within the next day.
[0049] S24. Input the electricity price over the most recent period into the electricity price prediction model to obtain the day-ahead electricity price prediction result.
[0050] Specifically, in this embodiment, the electricity prices of the most recent 3 days are input into the electricity price prediction model to obtain the electricity price prediction value for each sampling time in the next day. Then, the electricity price prediction value for each sampling time is denormalized to obtain the electricity price prediction result for the corresponding sampling time in the next day.
[0051] S25. Input the load power, electricity price, actual meteorological data and future short-term meteorological forecast data into the intraday short-term prediction model to obtain the short-term prediction result.
[0052] Specifically, in this embodiment, within a day, the load power, electricity price, actual meteorological data of the most recent 8 hours, and meteorological forecast data for the next 4 hours are input into the intraday short-term forecast model to obtain the load power forecast value and electricity price forecast value for each sampling time within the next 4 hours. Then, the load power forecast value and electricity price forecast value for each sampling time are denormalized to obtain the load forecast result and electricity price forecast result for the corresponding sampling time within the day. Both are used as the intraday short-term forecast result. The short-term forecast result is obtained every hour. In addition, in order to distinguish it from the load forecast result obtained in step S22 and the electricity price forecast result obtained in step S24, the load forecast result and electricity price forecast result obtained in this step are recorded as the short-term load forecast result and the short-term electricity price forecast result.
[0053] S3. Based on the day-ahead forecast results, and in conjunction with the rated capacity and state-of-charge operating range of the energy storage system, determine the control parameters for intraday regulation.
[0054] The control parameters include the benchmark electricity price, the electricity price sensitivity coefficient, the upper limit of the virtual state of charge (VFC), and the lower limit of the VFC. Step S3 specifically includes the following steps: S31. The load forecast results are summed and normalized, and the normalized load forecast results are used as the weights of the electricity price at the corresponding time. The weighted sum of the electricity price forecast results within the day is used as the benchmark electricity price.
[0055] Specifically, in this embodiment, for any sampling time within a future day, the weight of the electricity price at that time can be calculated using the following formula:
[0056] in, The weight of the electricity price at the j-th sampling time within the next day. This represents the load forecast result for the j-th sampling time within the next day.
[0057] Furthermore, the benchmark electricity price for the next day is calculated according to the following formula:
[0058] in, The benchmark electricity price, This represents the electricity price forecast for the j-th sampling time within the next day.
[0059] S32. Calculate the maximum daily electricity price difference based on the electricity price forecast results, and then calculate the ratio of the rated capacity to the maximum daily electricity price difference, and determine the electricity price sensitivity coefficient in combination with the safety margin coefficient.
[0060] Specifically, in this embodiment, the electricity price sensitivity coefficient satisfies the following relationship:
[0061] in, Electricity price sensitivity coefficient, For safety margin coefficient, The rated capacity of the energy storage system in the industrial park, The highest electricity price predicted for the next day. This represents the lowest electricity price predicted for the next day. The safety margin coefficient ranges from [0.7, 0.8], with 0.8 being the preferred value. This prevents the virtual state of charge calculated later from exceeding the limit under extreme electricity prices and can be adjusted manually.
[0062] S33. Update the benchmark electricity price and the electricity price sensitivity coefficient using the short-term forecast results according to the preset time step.
[0063] Specifically, in this embodiment, the preset time step is 1 hour. Every hour, referring to the method of calculating the benchmark electricity price in step S31 and the method of calculating the electricity price sensitivity coefficient in step S32, the benchmark electricity price and the electricity price sensitivity coefficient are recalculated using the short-term load forecast results and the short-term electricity price forecast results, thereby realizing the rolling update of the benchmark electricity price and the electricity price sensitivity coefficient. The recalculated electricity price sensitivity coefficient directly replaces the old electricity price sensitivity coefficient. However, when updating the benchmark electricity price, the recalculated benchmark electricity price is not directly used to replace the old benchmark electricity price. Instead, a first-order inertial filter is used to gradually transition from the old benchmark electricity price to the new benchmark electricity price.
[0064] in, Let be the base electricity price at time t. The old benchmark electricity price, For smoothing coefficients, This is the recalculated benchmark electricity price. A smoothing factor of 0.25 is used, with the goal of making... Every 5 minutes Keeping it close to 25% avoids subsequent abrupt changes in the virtual state of charge, which could cause the charging power command to jump instantaneously and damage the energy storage converter.
[0065] S34. Determine the upper limit of the virtual state of charge and the lower limit of the virtual state of charge using the state of charge operating range of the energy storage system.
[0066] Specifically, in this embodiment, the state-of-charge (SOC) operating range of the energy storage system is represented as follows: , This represents the upper limit of the state of charge of the energy storage system. This is the state-of-charge limit for the energy storage system. Typically... , The virtual state of charge upper limit and the virtual state of charge lower limit are used respectively. and It means, and , .
[0067] S4. Using the control parameters, the electricity price prediction results, the state of charge of the energy storage system, and the short-term prediction results, generate the charging and discharging power command of the energy storage system.
[0068] The charge / discharge power command includes a charging power command and a discharging power command. Step S4 specifically includes the following steps: S41. Using the control parameters, the short-term prediction results, and the state of charge of the energy storage system, calculate the virtual state of charge of the energy storage system in real time.
[0069] Specifically, in this embodiment, the virtual state of charge of the energy storage system satisfies the following relationship:
[0070] in, Let be the virtual state of charge at time t. Let t be the state of charge of the energy storage system. This represents the short-term electricity price forecast at time t. The state of charge (SOC) of the energy storage system refers to the actual SOC of the energy storage system in the industrial park. Furthermore, if the industrial park can obtain the market-clearing electricity price at time t, then... The preferred method is to use the market-clearing electricity price at time t.
[0071] S42. When the virtual state of charge is greater than or equal to the upper limit of the virtual state of charge, the energy storage system discharges, and the discharge power is calculated in combination with the rated power of the energy storage converter to form a discharge power command.
[0072] Specifically, in this embodiment, when the virtual state of charge... Greater than or equal to the upper limit of virtual state of charge At that time, the energy storage system discharges, and the discharge power is calculated based on the rated power of the energy storage converter. The discharge power satisfies the following relationship:
[0073] in, For discharge power, This refers to the rated power of the energy storage converter. This indicates discharge. The energy management platform of the industrial park generates discharge power commands for the energy storage system based on the calculated discharge power, therefore... It also represents the discharge power command.
[0074] S43. When the virtual state of charge is less than or equal to the virtual state of charge limit, the energy storage system is charged, and the charging power is calculated in combination with the rated power of the energy storage converter to form a charging power command.
[0075] Specifically, in this embodiment, when the virtual state of charge... Less than or equal to the virtual state of charge limit At that time, the energy storage system is charged, and the charging power is calculated based on the rated power of the energy storage converter. The charging power satisfies the following relationship:
[0076] in, For charging power, Indicates charging. A constant is set to prevent the denominator from being 0. The industrial park's energy management platform generates charging power commands for the energy storage system based on calculated charging power. It also represents a charging power command.
[0077] When the virtual state of charge Greater than the virtual state of charge limit And less than the upper limit of the virtual state of charge. At this time, the energy storage system is in standby mode, and the charging and discharging power is 0.
[0078] S5. After performing closed-loop correction and safety limiting on the charging and discharging power command based on the actual output power of the energy storage converter and the state of charge of the energy storage system, the command is sent to the energy storage converter for execution.
[0079] Step S5 specifically includes the following steps: S51. Calculate the power deviation between the charging / discharging power command and the actual output power of the energy storage converter, and then perform closed-loop correction of the charging / discharging power command through PI control.
[0080] Specifically, in this embodiment, the energy management platform of the industrial park acquires the actual output power of the energy storage converter once per second, and calculates the power deviation by combining it with the charging and discharging power command obtained in step S4:
[0081] Where e is the power deviation, This refers to either a discharge power command or a charging power command. Positive representation , negative indicates ; This represents the actual output power of the energy storage converter. A positive value indicates discharge. A negative value indicates charging. A power deviation greater than 0 indicates insufficient actual discharge or excessive charging, requiring increased discharge or decreased charging; a power deviation less than 0 indicates excessive actual discharge or insufficient charging, requiring decreased discharge or increased charging; a power deviation equal to 0 indicates appropriate actual discharge or charging.
[0082] Furthermore, the power deviation is fed into an incremental proportional-integral (PI) controller to calculate the required power correction.
[0083] in, This is the power correction amount for discharging or charging. This is the proportionality coefficient. The integral coefficient is... The sampling period. The default value is usually 0.5~2, with 0.5 being the preferred value; The default value is usually 5 to 20, with 5 being the preferred value.
[0084] The discharge power command is corrected using the discharge power correction factor to obtain the corrected discharge power command; similarly, the charging power command is corrected using the charging power correction factor to obtain the corrected charging power command. The corrected charging and discharging power commands satisfy the following relationship:
[0085] in, This refers to the revised charge / discharge power command, i.e., the revised discharge power command or charging power command.
[0086] S52. Based on the state of charge of the energy storage system, the modified charging and discharging power command is safely limited to obtain the final power command and sent to the energy storage converter for execution.
[0087] Specifically, in this embodiment, the modified charge and discharge power commands are safety-limited based on the state of charge of the energy storage system to obtain the final power command, which is then sent to the energy storage converter for execution. The final power command includes the final discharge power command and the final charge power command. The final power command satisfies the following relationship:
[0088] in, For the final power command, The state of charge of the energy storage system. This refers to the hysteresis bandwidth. The preferred hysteresis bandwidth is... Setting the hysteresis bandwidth is to prevent frequent switching of the limiting state when the state of charge fluctuates near the threshold.
[0089] It should be noted that in some cases, the actions described in the specification can be performed in different orders and still achieve the desired results. In this embodiment, the order of steps is given only to make the embodiment clearer and easier to explain, and not to limit it.
[0090] In one optional embodiment, please refer to Figure 2To improve the practicality of this method and facilitate its promotion, this invention also provides an energy storage system operation optimization system that takes into account dynamic electricity price signals. The energy storage system operation optimization system that takes into account dynamic electricity price signals includes: a data acquisition device 1, a data output device 2, a processor 3, and a storage device 4. The storage device 4 includes a computer-readable storage medium that stores a computer program. The computer program includes program instructions. When the program instructions are executed by the processor 3, the processor 3 performs the contents described in steps S1 to S5.
[0091] In summary, this invention has at least the following advantages: This method converts dynamic electricity price signals into state-of-charge (SOC) offsets and directly generates charging / discharging decisions through virtual threshold comparisons. This eliminates the need for online solutions to complex optimization problems such as mixed-integer programming or quadratic programming, reducing computational resource requirements and allowing deployment on resource-constrained edge computing devices in industrial parks. Furthermore, this method achieves real-time linkage between the virtual SOC and the actual SOC, allowing electricity price changes to be immediately reflected in the virtual SOC. Charging / discharging decisions do not require waiting for optimization cycles, resulting in a fast response to dynamic electricity price signals and effectively capturing spot market prices. This method addresses fluctuation opportunities; by determining control parameters at the day-ahead scale, it ensures overall economic efficiency within the scheduling cycle; at the intraday scale, it dynamically adjusts based on ultra-short-term forecasts to adapt to source-load fluctuations; and at the real-time scale, it provides power correction to ensure control accuracy, achieving an effective balance between long-term planning and short-term adaptation. Furthermore, this method uses the actual state of charge as the basis for hard constraints on charge and discharge commands, ensuring that the energy storage system always operates within a safe operating range, effectively preventing overcharging and over-discharging, and guaranteeing battery safety and stable system operation. Finally, it provides a system compatible with the method, which not only enhances its practicality but also facilitates its widespread adoption.
[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.
Claims
1. A method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals, characterized in that, Includes the following steps: Acquire data on the load power, energy storage system operation status, and photovoltaic power generation of the industrial park, and simultaneously acquire electricity price and meteorological data to build a historical dataset; Based on the historical dataset, the day-ahead load forecast, photovoltaic power generation forecast, and electricity price forecast are obtained, as well as the short-term intraday load and electricity price forecasts. Based on the day-ahead forecast results, combined with the rated capacity and state-of-charge operating range of the energy storage system, the control parameters for intraday regulation are determined. Using the control parameters, the electricity price prediction results, the state of charge of the energy storage system, and the short-term prediction results, a charging and discharging power command for the energy storage system is generated. After performing closed-loop correction and safety limiting on the charging and discharging power command based on the actual output power of the energy storage converter and the state of charge of the energy storage system, the command is sent to the energy storage converter for execution.
2. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 1, characterized in that: The energy storage system operation status data includes the energy storage system's state of charge and the actual output power of the energy storage converter. The photovoltaic power generation data is the photovoltaic output power of the photovoltaic power generation system. The meteorological data includes actual meteorological data and meteorological forecast data, and the meteorological forecast data includes temperature forecast data, humidity forecast data, wind speed forecast data, and sunshine forecast data.
3. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 2, characterized in that, The process of acquiring load power, energy storage system operation status data, and photovoltaic power generation data of the industrial park, while simultaneously acquiring electricity price and meteorological data, and then constructing a historical dataset, includes the following steps: Real-time collection of raw load power, raw energy storage system operation status data, and raw photovoltaic power generation data of industrial parks, and simultaneous acquisition of raw electricity prices and raw meteorological data; Preprocessing is performed on the raw load power, the raw energy storage system operating status data, the raw photovoltaic power generation data, the raw electricity price, and the raw meteorological data. A historical dataset is constructed using the preprocessed load power, the energy storage system operating status data, the photovoltaic power generation data, the electricity price, and the weather forecast data.
4. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 3, characterized in that, The process of obtaining day-ahead load forecasts, photovoltaic power generation forecasts, and electricity price forecasts based on the historical dataset, as well as short-term intraday load and electricity price forecasts, includes the following steps: A prediction model is constructed based on the historical dataset. The prediction model includes a load prediction model, a photovoltaic power generation prediction model, an electricity price prediction model, and an intraday short-term prediction model. The load power in the recent period, the actual meteorological data, and the meteorological forecast data for the next day are input into the load prediction model to obtain the load prediction results at the day-ahead scale. The photovoltaic output power in the recent period, the actual meteorological data, and the meteorological forecast data for the next day are input into the photovoltaic power generation prediction model to obtain the photovoltaic power generation prediction results on a day-ahead scale. The electricity price over a recent period is input into the electricity price forecasting model to obtain the day-ahead electricity price forecast result; The load power, electricity price, actual meteorological data, and future short-term meteorological forecast data are input into the intraday short-term prediction model to obtain the short-term prediction results.
5. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 4, characterized in that, The construction of a prediction model based on the historical dataset, comprising a load forecasting model, a photovoltaic power generation forecasting model, an electricity price forecasting model, and an intraday short-term forecasting model, includes the following steps: Based on the aforementioned historical dataset, the first dataset, the second dataset, the third dataset, and the fourth dataset are constructed respectively. The LSTM was trained and validated using the first dataset, the second dataset, the third dataset, and the fourth dataset, respectively, to obtain the load forecasting model, the photovoltaic power generation forecasting model, the electricity price forecasting model, and the intraday short-term forecasting model.
6. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 1, characterized in that: The control parameters include the benchmark electricity price, the electricity price sensitivity coefficient, the upper limit of the virtual state of charge, and the lower limit of the virtual state of charge.
7. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 6, characterized in that, Based on the day-ahead forecast results, and combined with the rated capacity and state-of-charge operating range of the energy storage system, the control parameters for intraday regulation are determined, including the following steps: The load forecast results are summed and normalized, and the normalized load forecast results are used as the weights of the electricity price at the corresponding time. The weighted sum of the electricity price forecast results within the day is used as the benchmark electricity price. The maximum daily electricity price difference is calculated based on the electricity price forecast results, and then the ratio of the rated capacity to the maximum daily electricity price difference is calculated. The electricity price sensitivity coefficient is determined in combination with the safety margin coefficient. The benchmark electricity price and the electricity price sensitivity coefficient are updated using the short-term forecast results according to a preset time step. The virtual state of charge upper limit and the virtual state of charge lower limit are determined using the state of charge operating range of the energy storage system.
8. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 7, characterized in that, The step of generating a charging and discharging power command for the energy storage system using the control parameters, the electricity price forecast results, the state of charge of the energy storage system, and the short-term forecast results includes the following steps: Using the aforementioned control parameters, the short-term prediction results, and the energy storage system's state of charge, the virtual state of charge of the energy storage system is calculated in real time. When the virtual state of charge is greater than or equal to the upper limit of the virtual state of charge, the energy storage system discharges, and the discharge power is calculated in combination with the rated power of the energy storage converter to form a discharge power command; When the virtual state of charge is less than or equal to the virtual state of charge limit, the energy storage system charges, and the charging power is calculated in combination with the rated power of the energy storage converter to form a charging power command.
9. The method for optimizing the operation of an energy storage system taking into account dynamic electricity price signals according to claim 1, characterized in that, The step of performing closed-loop correction and safety limiting on the charging and discharging power command based on the actual output power of the energy storage converter and the state of charge of the energy storage system before sending it to the energy storage converter for execution includes the following steps: The power deviation between the charging and discharging power command and the actual output power of the energy storage converter is calculated, and then the charging and discharging power command is corrected in a closed loop through PI control. The modified charging and discharging power command is safely limited based on the state of charge of the energy storage system to obtain the final power command, which is then sent to the energy storage converter for execution.
10. An energy storage system operation optimization system taking into account dynamic electricity price signals, characterized in that, The energy storage system operation optimization system taking into account dynamic electricity price signals includes: a data acquisition device, a data output device, a processor, and a storage device. The storage device includes a computer-readable storage medium storing a computer program. The computer program includes program instructions, which, when executed by the processor, cause the processor to implement the energy storage system operation optimization method taking into account dynamic electricity price signals as described in any one of claims 1-9.