An Optimal Scheduling Method for Source-Grid-Load-Storage Integration Projects

Abstract

The present invention belongs to the technical field of power system operation and dispatching, and specifically relates to an optimized dispatching method for a source-network-load-storage integrated project, which includes the following steps: preprocess the historical data of the output power of renewable energy and the load in combination with the given data sequence period of the source-network-load-storage area, and respectively construct and train a seasonal autoregressive integrated moving average model and a grey prediction model GM(1,1) for prediction. Then, obtain the final prediction results of the future output power of renewable energy and the load by weighted averaging of the two prediction data; construct a source-network-load-storage collaborative dispatching model considering the inverter to provide auxiliary services, input the prediction data into the collaborative dispatching model to obtain an initial dispatching plan; finally, use an improved tree species optimization algorithm to dispatch the distribution network and energy optimization in the source-network-load-storage. The present invention optimizes the safe dispatching of the distribution network to minimize the total market transaction cost, thereby improving the economic efficiency of the source-network-load-storage integrated project.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system operation and dispatching, and in particular relates to an optimization dispatching method for a source-grid-load-storage integrated project. Background Art

[0002] At present, the power system field is facing many challenges and also ushering in new development opportunities. The global demand for renewable energy is rising rapidly, especially relying on a high proportion of renewable energy, such as wind and solar energy, which has become a key strategy to reduce greenhouse gas emissions and respond to climate change. This trend requires the power system to adapt to a more complex energy structure and unstable energy supply, and many parts of the world have taken this model as the main development direction. The volatility of these energy sources significantly increases the uncertainty of the system, complicates the operating environment of the power grid, increases the operating risk, and may trigger large-scale power outages. At the same time, the rapid development of power electronics technology provides great potential for the reliability and efficiency of the power system. The high degree of electronicization of the system means that more power transmission, distribution and conversion processes will be completed by electronic equipment and control systems, greatly enhancing the flexibility and rapid response capabilities of the power system. However, this transformation of energy structure and power electronics also brings new challenges, the most notable of which is the problem of maintaining power balance. Due to the volatility of renewable energy, the power system must effectively manage the balance of power supply and demand to ensure the stability and reliability of power supply.

[0003] In this context, the integration of power generation, grid, load and storage is particularly important. This comprehensive strategy effectively coordinates the interaction between power generation, grid, load and energy storage systems by optimizing the production, transmission, distribution and storage of energy, thereby improving the efficiency and flexibility of the entire system. The integration of power generation, grid, load and storage can balance the supply and demand of electricity under unstable power supply conditions through efficient energy storage solutions and intelligent load management, thereby significantly improving the stability of the power grid and preventing power outages. Traditional power system dispatching strategies are not suitable for this new model, and there is an urgent need to further develop the application of power generation, grid, load and storage integration through innovative theories and technologies to meet the complex challenges facing future power systems. Summary of the invention

[0004] In view of the technical problem that the above-mentioned traditional power system dispatching strategy is not sufficiently adapted to this new mode, the present invention provides an optimization dispatching method for source-grid-load-storage integrated projects.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] An optimization scheduling method for a source-grid-load-storage integration project includes the following steps:

[0007] S1. Collect historical data at a fixed data sequence period and preprocess the historical data set. The historical data set includes light intensity data, photovoltaic power data, wind power data, electric vehicle load data, ambient temperature data, ambient humidity data, wind speed, wind direction, atmospheric pressure, irradiance, cloud cover, and traffic flow of electric vehicles recorded by the wind-solar power station during the historical period.

[0008] S2. Input the preprocessed historical data into the SARIMA model and the grey prediction model GM(1,1). SARIMA calculates and analyzes parameters such as the mean, variance, and residual confidence of the real data and the predicted data, and trains the model with these to obtain its initial optimal parameters, and then predicts the renewable energy output power and load data for the future time period. When the error of the subsequent predicted data does not meet the requirements, m historical data are dynamically updated and the SARIMA model parameters are re-optimized to improve the data prediction accuracy affected by seasons. GM(1,1) realizes the data prediction of the renewable energy output power and load with large short-term fluctuations. The GM(1,1) model parameters can be re-optimized by analyzing the changes in the mean square error ratio and the small error probability. To improve the prediction accuracy affected by seasons and short-term climate, combine the outputs of the SARIMA model and the grey prediction model GM(1,1), and obtain the final output power and load of the future renewable energy through weighted average.

[0009] S3. Install a power electronic controller on the distributed generation equipment, construct a source-network-load-storage coordinated scheduling model considering the inverter providing auxiliary services, and input the renewable energy output power and load data into the source-network-load-storage coordinated scheduling model to obtain an initial scheduling plan.

[0010] S4. Construct a day-ahead clearing model, and input the source-network-load-storage coordinated scheduling results considering the inverter providing auxiliary services into the model to verify the economy of the scheduling plan.

[0011] S5. Combine the coordinated scheduling model and the day-ahead clearing model, and use an improved tree species optimization algorithm to optimize the scheduling to minimize the total market transaction cost and make the distribution network meet the security check.

[0012] The method for preprocessing the historical data in S2 is as follows:

[0013] S2A. Denoise the light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles in the historical data through the wavelet transform algorithm.

[0014] S2B. Through the improved ensemble empirical mode decomposition algorithm, improve the accuracy of the decomposition results of the denoised light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles.

[0015] S2C. Normalize the algorithm result data to obtain the preprocessed data;

[0016] S2D. Normalize the environmental temperature data, environmental humidity data, atmospheric pressure, irradiance, and cloud amount of the historical data to obtain the preprocessed data.

[0017] The method for noise reduction of the light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles in the historical data by using the wavelet transform algorithm in S21 is as follows:

[0018] S2A1. Perform wavelet transform on the light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles. The wavelet transform formula is:

[0019]

[0020] where f is the data to be noise-reduced; a and b represent the scaling factor and the translation factor respectively; ψ(t) is the basic wavelet;

[0021] S2A2. Process using the soft threshold filtering method, select the optimal threshold through the general threshold and retain or correct the wavelet coefficients based on the threshold. The soft threshold filtering processing formula is:

[0022]

[0023] where f n is the wavelet coefficient; ε is the optimal threshold;

[0024] The calculation formula for the optimal threshold is:

[0025]

[0026] where σ is the noise standard deviation and N is the sum of the wavelet coefficients;

[0027] S2A3. Obtain the noise-reduced light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles through the inverse wavelet transform. The inverse wavelet transform calculation formula is:

[0028] f′ = CWT ―1 [F[CWT(f)]]

[0029] where f′(t) is the noise-reduced data.

[0030] In S22, the MEEMD algorithm is used to improve the accuracy of the decomposition results of the noise-reduced light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles, including the following steps:

[0031] S2B1. Add white noise signals with a mean of 0 to the denoised light intensity data, wind speed data, wind direction data and electric vehicle traffic flow, respectively. The addition formula is:

[0032]

[0033] Among them, n i (t) is the added white noise signal; a i is the amplitude of the added noise signal, i=1,2,...,M, M is the logarithm of the added white noise;

[0034] S2B2, yes and Perform empirical mode decomposition to obtain the first-order intrinsic modal component sequence, and integrate it to obtain the component s(t);

[0035] S2B3, check the entropy value of s(t). If it is less than or equal to η, it is not an abnormal signal; if it is greater than η, it is an abnormal signal, and return to S221 until the IMF component s p (t) is not an abnormal signal; where η is taken as 0.55 to 0.6;

[0036] S2B4. Eliminate the decomposed p-1 components from the light intensity data, wind speed data, wind direction data and the traffic flow of electric vehicles, where the elimination formula is:

[0037]

[0038] S2B5. Perform EMD decomposition on the residual signal h(t), arrange all the obtained IMF components from high frequency to low frequency, and synthesize the processed signal H(t).

[0039] The SARIMA model and the grey prediction model GM (1,1) in S2 are constructed and the prediction method for parameter optimization is as follows:

[0040] S2a1, according to the processed signal H(t), with a fixed data sequence period T, where T can be a number of months or weeks, obtain pre-processed historical data;

[0041] S2a2. Calculate and analyze the trend and periodic changes of the output power and load data of renewable energy using autocorrelation function and partial autocorrelation function;

[0042] S2a3. Make the time - series data stationary through differential operations. Determine the non - seasonal differencing order \(d\) and seasonal differencing order \(D\) of the SARIMA model according to the data characteristics. Use ACF and PACF to calculate the orders of autoregression and moving average, determine the maximum lag order \(p\) of the non - seasonal part and the maximum lag order \(q\) of the moving average operator, as well as the maximum lag order \(P\) of the seasonal part and the maximum lag order \(Q\) of the periodic moving average operator;

[0043] S2a4. Based on the analysis in the previous steps, determine the initial parameter combinations \((p,d,q)\) and \((P,D,Q,T)\), where \(T\) is the period length;

[0044] S2a5. Use the Bayesian information criterion to select the best model parameter combination, and then diagnose the fitted model. Verify that the model fits well according to the residual confidence;

[0045] S2a6. Use the fitted SARIMA model for prediction. According to the mean absolute error, root mean square error between the predicted value and the true value, and the residual exceeding the threshold, return to S2a3 to train and optimize each parameter of the model;

[0046] S2a7. Since some historical data are extremely vulnerable to climate influence, if the prediction error of one of the data is large, introduce the most recent \(m\) historical data, delete the earliest \(m\) historical data, and return to S2a1 to train and optimize each parameter of the model again;

[0047] S2b1: Obtain a fixed number of historical data, generate an accumulated sequence, and then generate a sequence of equal - weight neighboring values through weighted neighboring values;

[0048] S2b2. Construct a first - order one - variable differential equation in white - noise form based on the time series, and use the least - squares method to fit and solve to obtain the development grey number and control grey number, thus obtaining the grey prediction model GM(1,1);

[0049] S2b3. Use historical data for prediction. By calculating the mean square error ratio, if the mean square error ratio does not meet the requirements, return to step S2b1 and increase the number of data sequences, and recalculate the development grey number and control grey number of the model;

[0050] S2b4. When new historical data are generated, dynamically update the historical data of the output power and load of renewable energy respectively;

[0051] S2c. Combine the outputs of the SARIMA model and the grey prediction model GM(1,1), and calculate the final future output power and load of renewable energy through weighted averaging.

[0052] When considering the way of the inverter providing ancillary services in S3, the operating state of distributed generation is regulated by a droop - control model, and the specific formula is as follows:

[0053]

[0054] Among them, the active and reactive power output by the inverter in the distributed generator are set as P G,i and Q G,i , the output voltage phase angle of the inverter during stable operation is set to δ; E is the amplitude of the inverter output voltage, V is the node voltage amplitude of the distributed generation grid-connected node; Z is the self-impedance of the distributed generation;

[0055] The active power output of the inverter is proportional to the phase of the voltage, and the reactive power output is proportional to the voltage amplitude. The output voltage can be regulated by calculating the active and reactive power output of the inverter in each distributed power generation.

[0056] The method in which the inverter provides support for the safe and stable operation of the power grid and provides paid auxiliary services is:

[0057] S31, obtaining the three-phase voltage Vabc on the inverter side, the three-phase voltage Uabc on the grid side, and the three-phase current Iabc;

[0058] S32, coordinate conversion is performed on the three-phase voltage Vabc on the inverter side, the three-phase voltage Uabc on the grid side, and the three-phase current Iabc;

[0059] S33, inputting the three-phase voltage on the inverter side and the three-phase voltage on the grid side after coordinate conversion into a phase-locked loop module, obtaining the phase angle on the inverter side and the phase angle on the grid side, and calculating the power angle difference according to the phase angle on the inverter side and the phase angle on the grid side;

[0060] S34, using a PID controller to calculate the phase compensation angle of the power angle according to the power angle difference, and compensate the power angle according to the phase compensation angle; wherein the formula for calculating the phase compensation angle of the power angle using the PID controller is,

[0061] δ=K θp (θ ref ―θ v +θ u )+K θi ∫(θ ref ―θ v +θ u )dt

[0062] Among them, K θp is the phase compensation angle proportional gain, θ ref is the reference phase angle, θ v is the phase angle on the inverter side, θ u is the phase angle on the grid side, K θi is the phase compensation angle integral gain;

[0063] S35. Output an inverter control signal based on the compensated signal, control the inverter, and restore the power system fault.

[0064] The phase-locked loop module consists of a PI controller and an integral operation module. The PI controller adjusts the phase of the input voltage, and then inputs the adjusted voltage into the integral operation module to obtain the phase angle on the inverter side and the phase angle on the grid side. Among them, the control equation of the PI controller is:

[0065]

[0066] Among them, PI v is the output of the PI controller on the inverter side, K pv is the proportional gain of the voltage loop PI controller, K iv is the integral gain, K P is the proportional parameter, K i is the integral parameter, PI μ is the output of the PI controller on the grid side.

[0067] The objective function established for the day-ahead clearing model in S4 is:

[0068]

[0069] Among them, C is the total market transaction cost, Power is the output power of the source-grid-load-storage model, Load is the load of the source-grid-load-storage model, C sell is the revenue from selling electricity in the day-ahead market, C dh is the cost of purchasing electricity in the day-ahead market, C rm is the reserve capacity cost;

[0070] The cost of purchasing electricity in the day-ahead market is the cost of purchasing electricity from the external grid when the net output power of the source-grid-load-storage model is negative, including the cost of purchasing electricity from conventional power generators and the cost of purchasing electricity from new energy units; in the model, the conventional power generators are thermal power units, and the new energy units include wind power merchants and photovoltaic merchants; therefore, the cost of purchasing electricity in the day-ahead market can be expressed as:

[0071]

[0072] Among them, N G and N C are the number of conventional power generators and the number of quotes per period of each conventional power generator respectively, N W is the number of wind turbines, N PV is the number of photovoltaic units, ρ c,g,t is the market quote of conventional power generator g in the c-th segment at time t, P c,g,t is the cleared power of conventional power generator g in the c-th segment at time t, and They are the market quotation and cleared power of wind power merchant w during period t, and They are the market quotation and cleared power of photovoltaic power merchant pv during period t;

[0073] The revenue from selling electricity in the day-ahead market is the revenue from selling electricity to the external grid when the net output power of the source-grid-load-storage model is positive; The day-ahead market revenue can be expressed as:

[0074]

[0075] where ρ out,t is the market quotation of the external grid to the source-grid-load-storage model during period t; P out,t is the cleared power of the external grid to the source-grid-load-storage model during period t;

[0076] The reserve capacity cost is the cost incurred by the market to purchase reserve capacity from conventional units to reduce the adverse impact of the uncertainty of new energy output on the system; The reserve capacity cost includes two parts: the upward reserve capacity cost and the downward reserve capacity cost, and the mathematical expression is:

[0077]

[0078] where N g is the number of conventional units, are the upward reserve price and downward reserve price declared by conventional unit g during period t, respectively, are the upward reserve capacity and downward reserve capacity won by conventional unit g in the reserve capacity market during period t, respectively.

[0079] The specific steps of optimizing the dispatching of the distribution network and energy storage by using the improved tree species optimization algorithm based on the coordinated dispatching model and the day-ahead clearing model in S5 are as follows:

[0080] S51. Construct the adaptive search strategy of the improved tree species optimization algorithm as:

[0081]

[0082] where ST represents the adaptive search strategy, k represents the current iteration number, and k max is the maximum iteration number;

[0083] S52. Construct the mutation factor α of the improved tree species optimization algorithm as:

[0084] α = N(μ, σ 2 )

[0085] where μ is the midpoint between the individual optimal position and the global optimal position, and σ 2 is the deviation between the individual optimal position and the global optimal position;

[0086] S53. The expression of the improved tree species optimization algorithm is as follows:

[0087]

[0088] r = -1 + 2 * r1

[0089] Where X is the position of the current individual, r1 is a random number within the range of [0, 1], X best represents the optimal value of the current individual, X rand represents the position of a randomly selected individual within the population, and N represents the introduced mutation factor;

[0090] S54. Initialize the parameters of the improved tree species optimization algorithm; that is, take the active power output, reactive power output, and node voltage of the renewable energy generators in the scheduling model as the population;

[0091] S55. Obtain the optimal tree species of the current population, perform optimal scheduling based on this tree species, and perform safety verification on the scheduling result. If the verification passes, the optimal distribution network scheduling plan is obtained; otherwise, return to S54.

[0092] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0093] The present invention performs double noise reduction on the collected historical data and dynamically updates part of the historical data of renewable energy power generation and load data, so as to obtain accurate prediction data. Then, an adaptive search strategy, a mutation factor, and a balance mechanism are used to optimize the tree species algorithm, and then a source-network-load-storage collaborative scheduling model is constructed. Finally, considering that the inverter provides auxiliary services, a source-network-load-storage collaborative scheduling model is constructed, a scheduling plan is generated according to the scheduling model, and the improved tree species optimization algorithm is used to optimize the scheduling plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, without creative efforts, other implementation drawings can be obtained according to the provided drawings.

[0095] The structures, ratios, sizes, etc. shown in this specification are only used to cooperate with the content disclosed in the specification for those familiar with this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have a substantial technical meaning. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed in the present invention.

[0096] Figure 1 This is the overall network structure diagram of the present invention.

[0097] Figure 2 This is the double noise reduction flow chart of the historical data of the present invention.

[0098] Figure 3 This is the prediction flow chart of the output power and load of renewable energy of the present invention. Detailed implementation manners

[0099] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Apparently, the described embodiments are only a part rather than all of the embodiments of the present application. These descriptions are only for further explaining the features and advantages of the present invention, rather than limiting the claims of the present invention; based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present application.

[0100] The following will further describe in detail the specific implementation manners of the present invention with reference to the drawings and embodiments. The following embodiments are used to illustrate the present invention but not to limit the scope of the present invention.

[0101] An optimized scheduling method for a source-grid-load-storage integrated project, as Figure 1 shown, the method includes:

[0102] Step 1: Collect historical data in a fixed data sequence period and preprocess the historical data set, wherein the historical data set includes light intensity data, photovoltaic power data, wind power data, electric vehicle load data, environmental temperature data, environmental humidity data, wind speed, wind direction, atmospheric pressure, irradiance, cloud amount and traffic flow of electric vehicles recorded by a wind-solar power station in a historical period.

[0103] In this embodiment, the specific steps of Step 1 are as follows:

[0104] Step 11: Perform noise reduction processing on the light intensity data, wind speed data, wind direction data and traffic flow of electric vehicles in the historical data through a wavelet transform algorithm;

[0105] Step 111: Perform wavelet transform on the light intensity data, wind speed data, wind direction data and traffic flow of electric vehicles, where the wavelet transform formula is:

[0106]

[0107] Among them, f is the data to be denoised; a and b respectively represent the dilation factor and the translation factor; ψ(t) is the basic wavelet.

[0108] Step 112: Process using the soft threshold filtering method, select the optimal threshold through the general threshold and retain or correct the wavelet coefficients based on the threshold. The soft threshold filtering processing formula is:

[0109]

[0110] Among them, f n is the wavelet coefficient; ε is the optimal threshold.

[0111] The calculation formula for the optimal threshold is:

[0112]

[0113] Among them, σ is the noise standard deviation, and N is the sum of the wavelet coefficients.

[0114] Step 113: Obtain the denoised light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles through the inverse wavelet transform. The inverse wavelet transform calculation formula is:

[0115] f′ = CWT ―1 [F[CWT(f)]]

[0116] Among them, f′(t) is the denoised data.

[0117] Step 12: Improve the accuracy of the decomposition results of the denoised light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles through the improved ensemble empirical mode decomposition (MEEMD) algorithm;

[0118] Step 121: Add white noise signals with a mean of 0 to the denoised light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles respectively. The addition formula is:

[0119]

[0120] Among them, n i (t) is the added white noise signal; a i is the amplitude of the added noise signal, i = 1, 2,..., M, and M is the number of pairs of added white noise.

[0121] Step 122: Perform EMD decomposition on and to obtain the first-order IMF component sequence, and integrate them to obtain the component s(t).

[0122] Step 123: Check the entropy value of s(t). If it is less than or equal to η, it is not an abnormal signal; if it is greater than η, it is an abnormal signal, and return to Step 121 until the IMF component s p (t) is not an abnormal signal. Usually, η is taken as 0.55 - 0.6.

[0123] Step 124: Remove the p - 1 decomposed components from the light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles. The removal formula is:

[0124]

[0125] Step 125: Perform EMD decomposition on the remaining signal h(t), arrange all the obtained IMF components from high - frequency to low - frequency, and synthesize the processed signal H(t).

[0126] The double - noise reduction processing process for the light intensity data, wind speed data, wind direction data, and traffic flow of electric vehicles is as Figure 2 shown.

[0127] Step 13: Normalize the algorithm result data to obtain the pre - processed data;

[0128] Step 14: Normalize the environmental temperature data, environmental humidity data, atmospheric pressure, irradiance, and cloud amount of the historical data to obtain the pre - processed data.

[0129] Step 2: Input the pre - processed historical data into the SARIMA (Seasonal Autoregressive Integrated Moving Average) model and the grey prediction model GM(1,1). SARIMA calculates and analyzes parameters such as the mean, variance, and residual confidence of the real data and the predicted data, and trains the model with these to obtain its initial optimal parameters, and then predicts the output power and load data of renewable energy in the future time period. When the error of the subsequent predicted data does not meet the requirements, dynamically update m historical data and re - optimize the SARIMA model parameters to improve the prediction accuracy of the data affected by seasons. The GM(1,1) realizes the data prediction of the output power and load of renewable energy with large short - term fluctuations. By analyzing the changes in the mean square error ratio and the small error probability, the GM(1,1) model parameters can be re - optimized. To improve the prediction accuracy affected by seasons and short - term climate, combine the outputs of the SARIMA model and the grey prediction model GM(1,1), and obtain the final output power and load of renewable energy in the future through weighted average;

[0130] In this embodiment, the specific steps of Step 2 are as follows:

[0131] Step 211: Obtain historical data according to the processed signal H(t) with a fixed data sequence period T (T can be the number of months or weeks).

[0132] Step 212: Use the Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) to calculate and analyze the trends and periodic changes of renewable energy output power and load data.

[0133] Step 213: Make the time series data stationary through differencing operations, and determine the non-seasonal differencing order d and seasonal differencing order D of the SARIMA model according to the data characteristics.

[0134] Step 214: Use ACF and PACF to calculate the orders of autoregressive (AR) and moving average (MA), determine the maximum lag order p of the non-seasonal part, the maximum lag order q of the moving average operator, the maximum lag order P of the seasonal part, and the maximum lag order Q of the periodic moving average operator.

[0135] Step 215: Based on the analysis in the previous steps, determine the possible parameter combinations (p, d, q) and (P, D, Q, T), where T is the period length, and the SARIMA model can be expressed as:

[0136]

[0137] where B is the backshift operator, Δ d =(1 - B) d is the differencing operation, is the seasonal differencing operator, ε t is the white noise sequence; Φ p (B)=1 - φ1B 1 -…- φ p B p is the p-order autoregressive coefficient polynomial, Φ P (B T )=1 - a1B T -…- a P B PT is the P-order seasonal autoregressive parameter, Θ q (B)=1 - θ1B 1 -…- θ q B q is the q-order moving average coefficient polynomial, θ is the q-order moving average operator, Θ Q (B T )=1 - b1B T -…- bq B QT is the seasonal moving average parameter of order Q.

[0138] Step 216: Use the Bayesian Information Criterion (BIC) to select the best combination of model parameters, then diagnose the fitted model, and verify that the model fits well according to the residual confidence level;

[0139] BIC = K ln(n) - 2ln(L)

[0140] where K is the number of model parameters, n is the number of samples, and L is the maximum likelihood estimate of the model.

[0141] Step 217: Use the fitted SARIMA model to predict the output power and load of renewable energy, and evaluate the prediction accuracy of the model through the Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and residuals. When the evaluation index exceeds the threshold, return to Step 212 to train and optimize each parameter of the model.

[0142]

[0143] where n is the number of output power of renewable energy and distributed energy storage data, x v (i) is the true value, is the predicted value, and v takes 1 and 2 respectively to represent renewable energy and load.

[0144] Step 218: Since some historical data are extremely vulnerable to climate influence, if the prediction error of one of the data is large, introduce the most recent m (0 < m ≤ n) historical data, delete the earliest m historical data, and return to Step 211 to train and optimize each parameter of the model again.

[0145] Step 221: Obtain a fixed number of historical data sequences of the output power of renewable energy and load v = 1 represents the renewable energy type, v = 2 represents the load type, and generate the cumulative sequence Then generate the sequence of equal-weight neighboring values through weighted neighboring values

[0146] where

[0147] Step 222: Construct a first-order one-variable differential equation in white form based on the time series, and use the least squares method to fit and solve to obtain the development grey number α and the control grey number β, so as to obtain the grey prediction model GM(1,1) as follows:

[0148]

[0149] Step 223: Use historical data for prediction. By calculating the mean square error ratio, if the mean square error ratio does not meet the requirements, return to Step 221, increase the number of data sequences, and recalculate the development grey number and control grey number of the model.

[0150] Step 224: When new historical data is generated, dynamically update the historical data of the renewable energy output power and load respectively.

[0151] Step 23: To improve the prediction accuracy affected by seasons and short-term climate, combine the outputs of the SARIMA model and the grey prediction model GM(1,1), and obtain the final future output power and load of renewable energy through weighted average. The fused prediction data can be expressed as:

[0152]

[0153] Among them, x1 represents renewable energy, x2 represents load, and k represents data at the same time series. Therefore, when using the SARIMA model and the grey prediction model GM(1,1), it is necessary to ensure that the number of prediction data is the same.

[0154] The prediction process of the output power and load of renewable energy based on the combined SARIMA model and the grey prediction model GM(1,1) is as Figure 3 described.

[0155] Step 3: Construct a source-network-load-storage coordinated scheduling model considering the inverter providing auxiliary services, and input the predicted renewable energy output power and load data into the source-network-load-storage coordinated scheduling model to obtain an initial scheduling plan.

[0156] In this embodiment, the specific steps of Step 3 are as follows:

[0157] Step 31: Determine the output power according to the prediction result;

[0158] Step 32: When considering the inverter providing auxiliary services, the operating state of distributed generation is regulated by the droop control model, and the specific formula is as follows:

[0159]

[0160] Among them, the active and reactive powers output by the inverter in the distributed generator are set as P G,i and Q G,i , the output voltage phase angle of the inverter when it operates stably is set as δ; E is the amplitude of the inverter output voltage, V is the amplitude of the node voltage of the distributed generation grid-connected node; Z is the self-impedance of the distributed generation;

[0161] The active power output by the inverter is directly proportional to the phase of the voltage, and the reactive power output is directly proportional to the voltage amplitude. By calculating the active and reactive power outputs of the inverters in each distributed generation, the regulation of the output voltage is achieved.

[0162] Step 321: Obtain the three-phase voltages Vabc on the inverter side, the three-phase voltages Uabc on the grid side, and the three-phase current Iabc.

[0163] Step 322: Perform coordinate transformation on the three-phase voltages Vabc on the inverter side, the three-phase voltages Uabc on the grid side, and the three-phase current Iabc.

[0164] Step 323: Input the three-phase voltages on the inverter side and the three-phase voltages on the grid side after coordinate transformation into the phase-locked loop module to obtain the phase angles on the inverter side and the grid side, and calculate the power angle difference based on the phase angles on the inverter side and the grid side. The phase-locked loop module consists of a PI controller and an integral operation module. The PI controller adjusts the phase of the input voltage, and then inputs the adjusted voltage into the integral operation module to obtain the phase angles on the inverter side and the grid side. The control equation of the PI controller is:

[0165]

[0166] where, PI v is the output of the PI controller on the inverter side, K pv is the proportional gain of the voltage loop PI controller, K iv is the integral gain, K P is the proportional parameter, K i is the integral parameter, PI μ is the output of the PI controller on the grid side.

[0167] Step 324: Calculate the phase compensation angle of the power angle using a PID controller based on the power angle difference, and compensate the power angle according to the phase compensation angle. The formula for calculating the phase compensation angle of the power angle using a PID controller is:

[0168] δ = K θp (θ ref ― θ v + θ u ) + K θi ∫(θ ref ― θ v + θ u )dt

[0169] where, K θp is the proportional gain of the phase compensation angle, θ ref is the reference phase angle, θ vis the phase angle on the inverter side, θ u is the phase angle on the grid side, K θi is the integral gain of the phase compensation angle.

[0170] Step 325: Output an inverter control signal according to the compensated signal, and control the inverter to restore the power system fault.

[0171] Step 33: Obtain the energy consumption cost, demand-side response cost, energy storage loss cost, pollutant emission penalty cost, and the cost of the inverter providing services;

[0172] Calculate the function of the renewable energy consumption cost, expressed as:

[0173]

[0174] where C PV and C M are the cost electricity prices per unit output power of the photovoltaic and wind turbines respectively, N PV and N M are the numbers of the photovoltaic and wind turbines respectively, P PV,i,t and P M,i,t are the output powers of the photovoltaic and wind turbines respectively;

[0175] Calculate the function of the total cost of the thermal power unit, expressed as:

[0176]

[0177] where N F is the number of thermal power generating units, P it is the output of the thermal power unit i at time t, a it is the decision variable of the unit output, 0 represents not processing, 1 represents output, C s,it and C P,t are the energy consumption costs during unit startup and output respectively.

[0178] Calculate the function of the demand-side response cost, expressed as:

[0179]

[0180] where N DR is the number of responsive users, is the critical power of the response compensation price, P DRk,t and P DRl,t are the compensation amounts above and below respectively, ε a and ε b are the amounts by which P DR is higher and lower than Compensation coefficient at that time.

[0181] The function for calculating the energy storage loss cost is expressed as:

[0182]

[0183] Where N ES is the number of energy storage devices, C 0,it is the loss cost of energy storage device i at time t, C ES,i is the inherent investment of energy storage device i, P ES,i is the storage capacity of the device, is the discharge amount of the i-th energy storage device at time t, m i and n i are the charge and discharge coefficients.

[0184] The function for calculating the pollutant emission penalty cost is expressed as:

[0185]

[0186] Where N F is the number of thermal power generating units, λ ki , β ki , υ ki are all the relationship coefficients between the pollutant emissions and the output of the unit, ρ k is the relationship coefficient between the pollutant penalty cost and the emissions, s it is the influence factor of the pollutant emissions during the operation of the unit, generally between 0.8 and 1.3, P it is the output of thermal power generating unit i at time t.

[0187] The function for calculating the inverter service cost is expressed as:

[0188] F6 = C P P G,t + C Q Q G,t

[0189] Where C P and C Q are the compensation unit prices of the active power and reactive power of the virtual power plant respectively, P G,i and Q G,i are the active power and reactive power outputs of the inverter providing auxiliary services respectively.

[0190] Step 34: Construct an objective function based on the constraint conditions with the renewable energy generation cost, the energy consumption cost of the thermal power unit, the demand-side response cost, the energy storage loss cost, the pollutant emission penalty cost, and the inverter service cost;

[0191] The constraints include power balance constraints, capacity constraints of various components in the network, and operation constraints of various components in the network.

[0192] The power balance constraint is expressed as:

[0193] P load = P w + P PV + P f + P ES + P grid

[0194] where P load is the user load, P w is the wind power, P PV is the photovoltaic power, P f is the power of the thermal power generating unit, P ES is the power of the energy storage device (a positive value indicates discharging, and a negative value indicates charging), and P grid is the power transmitted from the power grid.

[0195] The capacity constraints of various components in the network include wind turbine capacity constraints, photovoltaic capacity constraints, thermal power generating unit capacity constraints, and energy storage device capacity constraints, and are expressed as:

[0196] The wind turbine capacity constraint is:

[0197] EW W,maxW,min

[0198] where E W,min and E W,max are the upper and lower limits of the wind turbine capacity respectively.

[0199] The photovoltaic capacity constraint is:

[0200] EPV PV,maxPV,min

[0201] where E PV,min and E PV,max are the upper and lower limits of the photovoltaic capacity respectively.

[0202] The thermal power generating unit capacity constraint is:

[0203] Ef f,maxf,min

[0204] where E f,min and E f,max are the upper and lower limits of the thermal power generating unit capacity respectively.

[0205] The energy storage device capacity constraint is:

[0206] EES ES,maxES,min

[0207] where EES,min , E ES,max are the upper and lower limits of the energy storage device capacity, respectively.

[0208] The operation constraints of each component in the network include wind power operation constraints, photovoltaic operation constraints, thermal power generation unit operation constraints, and energy storage device operation constraints, which are expressed as:

[0209] The wind power operation constraints are:

[0210] PW W,maxW,min

[0211] Among them, P W,min and P W,max are the upper and lower limit outputs of the wind turbine generator, respectively.

[0212] The photovoltaic operation constraints are:

[0213] PPV PV,maxPV,min

[0214] Among them, P PV,min and P PV,max are the upper and lower limit outputs of the photovoltaic unit.

[0215] The thermal power generation unit operation constraints are:

[0216] Pf f,maxf,min

[0217] Among them, P f,min and P f,max are the upper and lower limit outputs of the thermal power generation unit, respectively.

[0218] The energy storage device operation constraints are:

[0219]

[0220] Among them, S OC,min , S OC,max are the minimum and maximum state of charge of the energy storage device, respectively; are the maximum charging and discharging powers of the energy storage device, respectively.

[0221] Step 4: Construct a day-ahead clearing model and input the source-network-load-storage coordinated scheduling results of the distributed energy storage into the model to verify the economy of the scheduling scheme.

[0222] In this embodiment, the specific steps of Step 4 are as follows:

[0223] Step 41: Based on the source-network-load-storage coordinated scheduling results in the new power market environment, construct a day-ahead market clearing model including internal clearing and external clearing to verify the economic feasibility of the coordinated scheduling.

[0224] Step 42: Consider the joint optimization and clearing of multiple energy sources, with the minimum total market transaction cost as the optimization goal. The objective function consists of three parts: the revenue from selling electricity in the day-ahead market, the cost of purchasing electricity in the day-ahead market, and the reserve capacity cost.

[0225] The mathematical expression of the objective function of the day-ahead market clearing model is:

[0226]

[0227] Among them, C is the total market transaction cost, Power is the output power of the source-network-load-storage model, Load is the load of the source-network-load-storage model, C sell is the revenue from selling electricity in the day-ahead market, C dh is the cost of purchasing electricity in the day-ahead market,

[0228] C rm is the reserve capacity cost.

[0229] The cost of purchasing electricity in the day-ahead market is the cost of purchasing electricity from the external network when the net output power of the source-network-load-storage model is negative, including the cost of purchasing electricity from conventional power generators and the cost of purchasing electricity from new energy units. In the model, the conventional power generators are thermal power units, and the new energy units include wind power merchants and photovoltaic power merchants. Therefore, the cost of purchasing electricity in the day-ahead market can be expressed as:

[0230]

[0231] Among them, N G and N C are the number of conventional power generators and the number of quotes per period of each conventional power generator respectively, N W is the number of wind turbines, N PV is the number of photovoltaic units, ρ c,g,t is the market quote of conventional power generator g in the c-th segment at time t, P c,g,t is the cleared power of conventional power generator g in the c-th segment at time t, and are the market quote and cleared power of wind power merchant w at time t respectively, and are the market quote and cleared power of photovoltaic power merchant pv at time t respectively.

[0232] The revenue from selling electricity in the day-ahead market is the revenue from selling electricity to the external network when the net output power of the source-network-load-storage model is positive. The revenue in the day-ahead market can be expressed as:

[0233]

[0234] Among them, ρ out,t is the market quote of the external network to the source-network-load-storage model at time t; P out,t is the cleared power of the external network to the source-network-load-storage model at time t.

[0235] The reserve capacity cost is the cost incurred by the market for purchasing reserve capacity from conventional units to reduce the adverse impact of the uncertainty of new energy output on the system. The reserve capacity cost includes two parts: the upward reserve capacity cost and the downward reserve capacity cost. The mathematical expression is:

[0236]

[0237] Among them, N g is the number of conventional units, are the upward reserve price and the downward reserve price declared by conventional unit g in period t respectively, are the upward reserve capacity and the downward reserve capacity won by conventional unit g in the reserve capacity market in period t respectively.

[0238] Step 43: Set the constraint conditions; specifically including:

[0239] 1) System load balance constraint

[0240]

[0241] Among them, P t,g , P t,w , P t,pv are the outputs of conventional unit g, wind turbine w and PV power station pv in period t respectively; is the discharge of energy storage device i at time t; P c,g,t is the clearing power of conventional power generator g in section c of period t; are the clearing powers of wind power merchant w in period t respectively; are the clearing powers of PV power merchant pv in period t respectively; D t is the total system load in period t.

[0242] 2) Unit operation constraint

[0243]

[0244]

[0245] Among them, and are the minimum output power and the maximum output power limit of conventional power generator g respectively; is the maximum market declared power of wind power merchant w in period t; is the maximum market declared power of PV power merchant pv in period t; and are the upward and downward reserve capacities won by conventional unit g in the day-ahead market period t respectively; and They are the maximum up and down reserve capacities declared by the conventional unit g during the time period t, respectively.

[0246] 3) Network constraints

[0247]

[0248] P ij,t =(θ i,t ―θ j,t ) / x ij

[0249]

[0250] Where B is the imaginary part of the nodal admittance matrix; θ t is the column vector of nodal voltage phase angles at time t; is the column vector of active power injections at nodes at time t; P ij,t and are the branch active power and power limit of the first node i and the last node j at time t, respectively; θ i,t is the voltage phase angle of node i at time t; x ij is the reactance of branch ij.

[0251] 4) System reserve capacity constraint

[0252]

[0253] In the formula, N g is the number of conventional units, and are the maximum up reserve capacity and the maximum down reserve capacity that the conventional unit g can declare during the time period t, respectively, and are the system up reserve capacity and the system down reserve capacity during the time period t, respectively, and u t,g is the reserve capacity coefficient.

[0254] Step 5: Combine the coordinated dispatch model and the day-ahead clearing model, and use the improved tree seed optimizer algorithm (ITSA) to optimize the dispatch to minimize the total cost of market transactions and make the distribution network meet the security check, so as to obtain the optimal distribution network dispatch plan.

[0255] In this embodiment, the specific steps of Step 5 are as follows:

[0256] Step 51: Construct the adaptive search strategy of the improved tree seed optimizer algorithm (ITSA) as:

[0257]

[0258] where ST represents the adaptive search strategy, k represents the current iteration number, and k max is the maximum iteration number.

[0259] Step 52: Construct the mutation factor α of the improved tree species optimization algorithm as:

[0260] α = N(μ, σ 2 )

[0261] where μ is the midpoint between the individual optimal position and the global optimal position, and σ 2 is the deviation between the individual optimal position and the global optimal position.

[0262] Step 53: The expression of the improved tree species optimization algorithm is:

[0263]

[0264] r = -1 + 2 * r1

[0265] where X is the position of the current individual; r1 is a random number within the range of [0, 1], X best represents the current individual optimal value, and X rand represents the randomly selected individual position within the population, and N represents the introduced mutation factor.

[0266] Step 54: Initialize the parameters of the improved tree species optimization algorithm, that is, take the active power output, reactive power output of the renewable energy generator in the scheduling model, and the node voltage as the population;

[0267] Step 55: Obtain the optimal tree species of the current population, perform optimal scheduling based on this tree species, and conduct a security check on the scheduling result. If the check passes, the optimal distribution network scheduling plan is obtained; otherwise, return to Step 54.

[0268] The above only elaborates on the preferred embodiments of the present invention in detail. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, various changes can be made without departing from the purpose of the present invention, and all such changes should be included within the protection scope of the present invention.

Claims

1. An optimization scheduling method for a source-grid-load-storage integrated project, characterized in that: The following steps are involved: S1. Collect historical data at a fixed data sequence period and preprocess the historical data set, wherein the historical data set includes light intensity data, photovoltaic power data, wind power data, electric vehicle load data, ambient temperature data, ambient humidity data, wind speed, wind direction, atmospheric pressure, irradiance, cloud cover and electric vehicle traffic flow recorded by the wind and solar power station in the historical period; S2. Input the preprocessed historical data into the SARIMA model and the grey prediction model GM(1,1), where SARIMA calculates and analyzes the mean, variance, residual confidence and other parameters of the real data and the predicted data, and uses this to train the model to obtain its initial optimal parameters, and then predict the renewable energy output power and load data for the future time period. When the subsequent prediction data error does not meet the requirements, dynamically update the m historical data and re-optimize the SARIMA model parameters to improve the prediction accuracy of data affected by seasons, while GM(1,1) realizes the data prediction of renewable energy output power and load with large short-term fluctuations. The GM(1,1) model parameters can be re-optimized by analyzing the changes in the mean square error ratio and the probability of small errors. In order to improve the prediction accuracy affected by seasons and short-term climate, the outputs of the SARIMA model and the grey prediction model GM(1,1) are combined to obtain the final future output power and load of renewable energy through weighted averaging; S3, install power electronic controllers on distributed power generation equipment, build a source-grid-load-storage coordinated scheduling model that considers the inverter providing auxiliary services, input the renewable energy output power and load data into the source-grid-load-storage coordinated scheduling model, and obtain an initial scheduling plan; when considering the inverter providing auxiliary services in S3, the operating state of distributed power generation is regulated by the droop control model, and the specific formula is as follows: Among them, the active and reactive power output by the inverter in the distributed generator are set as P G,i and Q G,i , the output voltage phase angle of the inverter during stable operation is set to δ; E is the amplitude of the inverter output voltage, V is the node voltage amplitude of the distributed generation grid-connected node; Z is the self-impedance of the distributed generation; The active power output of the inverter is proportional to the phase of the voltage, and the reactive power output is proportional to the voltage amplitude. The output voltage can be regulated by calculating the active and reactive power output of each inverter in the distributed generation. S4. Build a day-ahead clearing model and input the source-grid-load-storage coordinated dispatch results that consider the inverter providing auxiliary services into the model to verify the economic feasibility of the dispatch scheme; S5. Combining the coordinated dispatch model and the day-ahead clearing model, an improved tree species optimization algorithm is used to optimize the dispatch to minimize the total market transaction cost and ensure that the distribution network meets the safety verification; The specific steps of optimizing the dispatching of distribution network and energy storage by using the improved tree species optimization algorithm based on the coordinated dispatching model and the day-ahead clearing model in S5 are as follows: S51. The adaptive search strategy for constructing the improved tree species optimization algorithm is: Among them, ST represents the adaptive search strategy, k represents the current number of iterations, and k max is the maximum number of iterations; S52. The variation factor α of the improved tree species optimization algorithm is constructed as follows: α=N(μ,σ 2 ) Among them, μ is the midpoint between the individual optimal position and the global optimal position, σ 2 is the deviation between the individual optimal position and the global optimal position; S53. The improved tree species optimization algorithm expression is: r＝-1+2*r1 Among them, X is the position of the current individual, r1 is a random number in the range of [0, 1], and X best represents the current individual optimal value, X rand represents the position of a random individual selected from the population, and N represents the introduced variation factor; S54, initializing the parameters of the improved tree species optimization algorithm; namely, taking the active output and reactive output of the renewable energy generators and the node voltage in the scheduling model as the population; S55. Obtain the optimal tree species of the current population, use this tree species to carry out optimized scheduling and perform safety verification on the scheduling results. If the verification passes, the optimal distribution network scheduling plan is obtained, otherwise return to S54.

2. The optimization scheduling method for source-grid-load-storage integrated projects according to claim 1 is characterized in that: The method for preprocessing historical data in S2 is: S2A, performing noise reduction processing on the light intensity data, wind speed data, wind direction data and traffic flow of electric vehicles of the historical data by using a wavelet transform algorithm; S2B, improve the accuracy of the decomposition results of the denoised light intensity data, wind speed data, wind direction data and electric vehicle traffic flow through the improved overall average empirical mode decomposition algorithm; S2C, normalize the algorithm result data to obtain preprocessed data; S2D. Normalize the ambient temperature data, ambient humidity data, atmospheric pressure, irradiance, and cloud cover of the historical data to obtain pre-processed data.

3. The optimization scheduling method for source-grid-load-storage integrated projects according to claim 2 is characterized in that: The method for performing noise reduction processing on the light intensity data, wind speed data, wind direction data and traffic flow of electric vehicles in the historical data by using the wavelet transform algorithm in S21 is: S2A1. Perform wavelet transform on the light intensity data, wind speed data, wind direction data and the traffic flow of electric vehicles, where the wavelet transform formula is: Where f is the data to be denoised; a and b represent the scaling factor and translation factor respectively; ψ(t) is the basic wavelet; S2A2, using the soft threshold filtering method, select the optimal threshold through the general threshold and use the threshold as the standard to retain or modify the wavelet coefficients; the soft threshold filtering processing formula is: Among them, f n is the wavelet coefficient; ε is the optimal threshold; The formula for calculating the optimal threshold is: Where σ is the standard deviation of noise, N is the sum of wavelet coefficients; S2A3, the noise-reduced light intensity data, wind speed data, wind direction data and electric vehicle traffic flow are obtained by inverse wavelet transform, where the inverse wavelet transform calculation formula is: f′=CWT -1 [F[CWT(f)]] Among them, f′(t) is the data after noise reduction.

4. The optimization scheduling method for source-grid-load-storage integrated projects according to claim 2 is characterized in that: The MEEMD algorithm is used in S22 to improve the accuracy of the de-noised light intensity data, wind speed data, wind direction data and the traffic flow decomposition result of electric vehicles, including the following steps: S2B1. Add white noise signals with a mean of 0 to the denoised light intensity data, wind speed data, wind direction data and electric vehicle traffic flow, respectively. The addition formula is: f i + (t)=f′(t)+a i n i (t) f i - (t)=f′(t)-a i n i (t) Among them, n i (t) is the added white noise signal; a i is the amplitude of the added noise signal, i=1,2,...,M, M is the logarithm of the added white noise; S2B2, f i + (t) and f i - (t) Perform empirical mode decomposition to obtain the first-order intrinsic modal component sequence, and integrate it to obtain the component s(t); S2B3, check the entropy value of s(t). If it is less than or equal to η, it is not an abnormal signal; if it is greater than η, it is an abnormal signal, and return to S221 until the IMF component s p (t) is not an abnormal signal; where η is taken as 0.55 to 0.6; S2B4. Eliminate the decomposed p-1 components from the light intensity data, wind speed data, wind direction data and the traffic flow of electric vehicles, where the elimination formula is: S2B5. Perform EMD decomposition on the residual signal h(t), arrange all the obtained IMF components from high frequency to low frequency, and synthesize the processed signal H(t).

5. The optimization scheduling method for source-grid-load-storage integrated projects according to claim 1 is characterized in that: The SARIMA model and the grey prediction model GM (1,1) in S2 are constructed and the prediction method for parameter optimization is as follows: S2a1, according to the processed signal H(t), with a fixed data sequence period T, where T can be a number of months or weeks, obtain pre-processed historical data; S2a2. Calculate and analyze the trend and periodic changes of the output power and load data of renewable energy using autocorrelation function and partial autocorrelation function; S2a3, make the time series data stable through difference operation, determine the non-seasonal difference order d and seasonal difference order D of the SARIMA model according to the data characteristics, use ACF and PACF to calculate the order of autoregression and moving average, determine the maximum lag order p of the non-seasonal part and the maximum lag order q of the moving average operator, as well as the maximum lag order P of the seasonal part and the maximum lag order Q of the periodic moving average operator; S2a4, based on the analysis in the previous step, determine the initial parameter combination (p, d, q) and (P, D, Q, T), where T is the cycle length; S2a5, use the Bayesian information criterion to select the best combination of model parameters, then diagnose the fitted model and verify that the model fits well based on the residual confidence level; S2a6, use the fitted SARIMA model to make predictions, and return to S2a3 to train and optimize the model parameters when the mean absolute error, root mean square error and residual between the predicted value and the true value exceed the threshold; S2a7, since some historical data are very easily affected by climate, if there is a large prediction error in one of the data, the most recent m historical data are introduced, and the earlier m historical data are deleted, and S2a1 is returned again to train and optimize the model parameters; S2b1: Obtain a fixed number of historical data, generate a cumulative sequence, and then generate an equal-weighted neighbor value generation sequence by weighted neighbor values; S2b2, construct a first-order one-variable differential equation based on the whitening form of time series, and use the least squares method to fit and solve the development gray number and control gray number, thereby obtaining the gray prediction model GM (1,1); S2b3, using historical data to make predictions, by calculating the mean square error ratio, if the mean square error ratio does not meet the requirements, return to step S2b1, increase the number of data sequences, and recalculate the development gray number and control gray number of the model; S2b4. When new historical data is generated, the historical data of the output power and load of renewable energy are dynamically updated respectively; S2c, combining the outputs of the SARIMA model and the grey prediction model GM(1,1), calculates the final future output power and load of renewable energy by weighted average.

6. The optimization scheduling method for source-grid-load-storage integrated projects according to claim 1 is characterized in that: The method in which the inverter provides support for the safe and stable operation of the power grid and provides paid auxiliary services is: S31, obtaining the three-phase voltage Vabc on the inverter side, the three-phase voltage Uabc on the grid side, and the three-phase current Iabc; S32, coordinate conversion is performed on the three-phase voltage Vabc on the inverter side, the three-phase voltage Uabc on the grid side, and the three-phase current Iabc; S33, inputting the three-phase voltage on the inverter side and the three-phase voltage on the grid side after coordinate conversion into a phase-locked loop module, obtaining the phase angle on the inverter side and the phase angle on the grid side, and calculating the power angle difference according to the phase angle on the inverter side and the phase angle on the grid side; S34, using a PID controller to calculate the phase compensation angle of the power angle according to the power angle difference, and compensate the power angle according to the phase compensation angle; wherein the formula for calculating the phase compensation angle of the power angle using the PID controller is, δ=K θp (i ref -θ v +θ u )+K θi ∫(θ ref -θ v +θ u )dt Among them, K θp is the phase compensation angle proportional gain, θ ref is the reference phase angle, θ v is the phase angle on the inverter side, θ u is the phase angle on the grid side, K θi is the phase compensation angle integral gain; S35. Output an inverter control signal according to the compensated signal, and control the inverter to restore the power system fault.

7. The optimization scheduling method for source-grid-load-storage integrated projects according to claim 6 is characterized in that: The phase-locked loop module is composed of a PI controller and an integral operation module. The PI controller is used to adjust the phase of the input voltage, and the adjusted voltage is input into the integral operation module to obtain the phase angle of the inverter side and the phase angle of the grid side. The control equation of the PI controller is: Among them, PI v is the output of the inverter side PI controller, K pv is the proportional gain of the voltage loop PI controller, K iv is the integral gain, K P is the scale parameter, K i is the integral parameter, PI μ is the output of the grid-side PI controller.

8. The optimization scheduling method for source-grid-load-storage integrated project according to claim 1 is characterized in that: The objective function established for the day-ahead clearing model in S4 is: Among them, C is the total market transaction cost, Power is the output power of the source-grid-load-storage model, Load is the load of the source-grid-load-storage model, and C sell is the revenue from electricity sales in the day-ahead market, C dh is the day-ahead market electricity purchase cost, C rm It is the reserve capacity fee; The day-ahead market electricity purchase cost is the electricity purchase cost from the external grid when the net output power of the source-grid-load-storage model is negative, including the electricity purchase cost of conventional power generators and the electricity purchase cost of new energy units. In the model, conventional power generators are thermal power units, and new energy units include wind power generators and photovoltaic power generators. Therefore, the day-ahead market electricity purchase cost can be expressed as: Among them, N G and N C are the number of conventional power generators and the number of quotations made by conventional power generators in each period, N W is the number of wind turbines, N PV is the number of photovoltaic units, ρ c,g,t is the market price of conventional power producer g in period c during period t, P c,g,t is the power cleared by conventional power generator g in the cth segment of period t, and are the market price and clearing power of wind power supplier w in period t, and are the market quotation and clearing power of photovoltaic electricity supplier pv in period t respectively; The revenue from electricity sales in the day-ahead market is the revenue from selling electricity to the external grid when the net output power of the source-grid-load-storage model is positive. The day-ahead market revenue can be expressed as: Among them, ρ out,t is the market quotation of the external network to the source network load storage model in period t; out,t is the clearing power from the external grid to the source grid-load-storage model in period t; The reserve capacity fee is the cost incurred by the market to purchase reserve capacity from conventional units in order to reduce the adverse impact of uncertainty in renewable energy output on the system. The reserve capacity fee includes upper reserve capacity fee and lower reserve capacity fee. The mathematical expression is: Among them, N g is the number of conventional units, are the upper reserve price and lower reserve price declared by the conventional unit g in period t, They are the upper and lower reserve capacities bid by the conventional unit g in the reserve capacity market during period t respectively.

Images