In this paper the costs of integrating fluctuating sources like wind or solar into an existing electricity system are quantified using a systematic, theoretically well-founded approach. The paper notably stresses that this requires to measure integration costs against some reference technology and it highlights the link between integration costs and changes in system costs. The costs related to wind and solar integration are furthermore decomposed into several components, including notably the costs of variability and the cost of unpredictability. It is then discussed, how a rough assessment of these cost components may be obtained using peak-load pricing concepts. But it is then highlighted that a detailed assessment requires stochastic optimization approaches.